Sei sulla pagina 1di 37

Beyond Credit Risk:

Economic Capital Models


− News and Views −

March 2006

G
GRC5MP, March 2006
Seite 1 12345
Draft Outline

• Introduction to Economic Capital Models

• The Problem of Hard-To-Measure Risks


A new approach for quantifying Strategic Risk

G
GRC5MP, March 2006
Seite 2 12345
Economic Capital Model: The Consultancy's View

G
GRC5MP, March 2006
Seite 3 12345
Four questions...

What is Philosophy?
(Immanuel Kant, 18th century)

What can we know ?

What should we do ?

What may we hope for ?

What is a human being ?

G
GRC5MP, March 2006
Seite 4 12345
Four questions concerning Economic Capital

What is Philosophy ? What is an EC Model ?


(Immanuel Kant, 18th century) (K.B. 21th century)

What is a human being ? What is EC ?

What can we know ? What can we measure ?

What should we do ? What should we do ?

What may we hope for ? What may we hope for ?

G
GRC5MP, March 2006
Seite 5 12345
What is Economic Capital?

G
GRC5MP, March 2006
Seite 6 12345
What is Economic Capital?
Questions…

Suppose you are management board member of a bank...

• How do we choose between


- an increase of EUR 20 mn in 1-day market VaR
- an increase of EUR 10 bn of exposure in private customers?

• What effect does an acquisition have on the overall risk profile


of our bank?

• Does our bank have enough capital to sustain potentially


extreme losses with sufficient high probability?

G
GRC5MP, March 2006
Seite 7 12345
What is Economic Capital?
…and the answer...

• We need a uniform risk measure across risk types,


business lines, regions, products, etc.
Economic Capital
• Here, risk is the unexpected loss of economic value over a certain
time period, calculated at very high confidence level

• Economic Capital measures risk from the debt holder perspective


(potential loss of value) rather than from the equity holder
perspective (volatility)

EC = Loss at very high CL (quantile) − Expected Loss (EL)

G
GRC5MP, March 2006
Seite 8 12345
What is Economic Capital?
Economic Capital at HVB
Economic Capital is based on a Value-at-Risk (VaR) with
• a time-horizon of one year
• a confidence level of 99.95 %.

Mean q99,95%

EL EC

covered losses against it


by net would be uneconomical
margin to hold capital
income

Loss
G
GRC5MP, March 2006
Seite 9 12345
What can we measure?

G
GRC5MP, March 2006
Seite 10 12345
What can we measure?
Economic Capital by risk types

Risk types amenable for quantification

market
• market risk 4%

• credit risk financial inv.


31%
• operational risk credit
34%
• business risk
• real estate risk real estate
4%
• financial investment risk
business oprisk
14% 13%

G
GRC5MP, March 2006
Seite 11 12345
What can we measure?
Aggregation of risk: Overview (1)
• Diversification is a competitive advance and should be
taken into account
- between business units
- between risk types.

• Dependence between different risk types is based on


realistic correlation estimates (InterCorrelation).

• Currently, the overall diversification benefit by InterCorrelation


is about 35 %.

G
GRC5MP, March 2006
Seite 12 12345
What can we measure?
Aggregation of risk: Overview (2)
1. Determine EC for each risk type separately, including
diversification effects between business units.
2. Aggregate those EC numbers into one EC for the entire bank.

Risk types EC per risk type Total EC


Business units

InterCorr.

G
GRC5MP, March 2006
Seite 13 12345
What can we measure?
InterCorrelation: Var-covar approach
• Correlations estimates between risk types are based on the
loss drivers as well as expert opinions.
- what drivers to use for time-varying positions in market risk?
- loss drivers for operational risk?

• Normal distribution: Quantiles can be expressed as multiples


of the standard deviation,

σ total = σ + σ + 2 ρ σ1σ 2
2
1
2
2

VaRtotal ≈ VaR12 + VaR22 + 2 ρVaR1VaR2

G
GRC5MP, March 2006
Seite 14 12345
What can we measure?
InterCorrelation: Copula approach
1. Choose appropriate marginal distributions for each risk type, e.g.
- market risk: normal distribution, Student t distribution
- credit risk: beta distribution, Vasicek distribution
- operational risk: lognormal distribution

2. Choose a parametric copula based on a correlation matrix

Gaussian Student t
tail dependence
copula copula

∞ degree of freedom 1

G
GRC5MP, March 2006
Seite 15 12345
What can we measure?
InterCorrelation: Total EC after portfolio effects

Total EC
+4%

-14 %

about
EUR 9 bn

G
GRC5MP, March 2006
Seite 16 12345
What should we do?

G
GRC5MP, March 2006
Seite 17 12345
What should we do?

Regulatory • Basel II

? requirements • MaRisk

Bank internal • Value based management


use of EC • Analysis of risk taking capacity

G
GRC5MP, March 2006
Seite 18 12345
What should we do?
Regulatory requirements

Basel II Framework

Pillar 1 Pillar 2 Pillar 3

Minimum capital requirements Supervisory Review Enhanced Disclosure


Process (SRP)

• quantitative and qualitative • emphasizes the need for a


guidelines. qualitative approach to
supervising banks.
- internal models for OpRisk
- no internal models for
credit risk!
•Internal Capital Adequacy
Assessment Process (ICAAP).

G
GRC5MP, March 2006
Seite 19 12345
What should we do?
Risk taking capacity
Does the bank have enough capital to support the risks it takes?

Risk Appetite:• other risks


?
What is enough?
• model risk Capital cushion basically consists of
• estimation errors • Net income forecast
• price and property reserves
• Hybrid capital instruments
• IFRS Shareholders‘ equity
minus Goodwill

Total EC Capital cushion


(incl. diversification)

G
GRC5MP, March 2006
Seite 20 12345
What may we hope for?

G
GRC5MP, March 2006
Seite 21 12345
What may we hope for?

• Better knowledge transfer from theory into practice

• Treatment of Hard-to-measure risks

G
GRC5MP, March 2006
Seite 22 12345
What may we hope for?
Theory-practice gap
There is a decrease of knowledge when moving from theory to practice.

Theory / Academics

"…a little knowledge is more successful


Knowledge than complete knowledge: it conceives
things as simpler than
they are, …[and is therefore]… more
comprehensible and persuasive."
[Friedrich Nietzsche, Human, all too human (1878)]

Practise / Industry

G
GRC5MP, March 2006
Seite 23 12345
What may we hope for?
Decrease of knowledge: an example
Little knowledge: Risk control Complete knowledge: Risk control
is a cost-intensive regulatory is not only a regulatory
requirement and diminishes requirement, but is an important
the bank's profit. contribution to success in banking.

little knowledge complete knowledge


profit distribution profit distribution
profit without RM
profit with RM

profit profit
0 E' E 0 E E'

G
GRC5MP, March 2006
Seite 24 12345
What may we hope for?

It would be a mistake to conclude that the only


way to succeed in banking is through
ever-greater size and diversity.
Indeed, better risk management may be the only truly
necessary element of success in banking.
Alan Greenspan,
Speach to the American Bankers Association, October 5, 2004

G
GRC5MP, March 2006
Seite 25 12345
What may we hope for?
Can a bank's total EC be quantified?

„Not everything that counts


can be counted...“
[Albert Einstein (1879 -1955)]

However:

„It is better to light one candle than


to curse the darkness.“
[unknown]

G
GRC5MP, March 2006
Seite 26 12345
What may we hope for?
Regulatory requirements

Basel II: "Although the Committee recognizes that ‘other’ risks […]
are not easily measurable, it expects industry to further develop
techniques for managing all aspects of these risks."

Qualitative Quantitative
Management by "intuition" "Light one candle" and try to at
in order to least approximate these risks.
• anticipate • VaR
• monitor • EC Model
• mitigate • Risk taking capacity
these risks.

G
GRC5MP, March 2006
Seite 27 12345
What may we hope for?
Strategy Evaluation
Corporate strategies often focus on expectations about future
earnings, which are based on the "most likely" future scenario.
earnings
(cash flows) lively rabbit

Present value:
Scenario ∞
∆ Ei
lame duck
P=∑
i =1 (1 + r ) i

t
0 1 2 3 4 5

"Convictions are more dangerous enemies of truth than lies"


[Friedrich Nietzsche]

G
GRC5MP, March 2006
Seite 28 12345
Good Strategy or Bad Strategy?

Ludwig Wittgenstein (1889-1951)


(1953/1958).

G
GRC5MP, March 2006
Seite 29 12345
What may we hope for?
Strategic risk: A question of measure

• The goal of every strategy is to generate cash flows

• A good strategy meets or exceeds expectations

• A bad strategy results in earnings below expectation,


thereby diminishing the market value of the company

The natural risk measure for strategic risk is VaR!

G
GRC5MP, March 2006
Seite 30 12345
What may we hope for?
A simple model for strategic risk

∆Ei
Recall the definition of present value: P=∑
i =1 (1 + r ) i

Definition: The present value in continuous time is given by


t
P (t ) = ∫ e − r ( s ) s dE ( s ), t ≥ 0,
0
and t
E (t ) = ∫ dE ( s )
0
is referred to as cumulated earnings at time horizon t.

G
GRC5MP, March 2006
Seite 31 12345
What may we hope for?
A simple model for strategic risk
To account for the uncertainty of future earnings, we introduce an
earnings process E (t ) t ≥given
0 by
dE (t ) = α (t ) dt + σ dW (t ), σ = const.
where α (t ) is deterministic and W (t ) is a 1-dim. Brownian motion.
earnings
35
α (t ) = 0.1
30
σ = 0.5
25

20

15

10

t
20 40 60 80 100

G
GRC5MP, March 2006
Seite 32 12345
What may we hope for?
A simple model for strategic risk
Let r be a constant discount rate and E (t ) t ≥0 the earnings process
given above. Then, the present value P (t ) is given by the value
process P (t ) t ≥0,
t
P (t ) = ∫ e − r s dE ( s ) = I1 (t ) + I 2 (t )
0
with t
I1 (t ) = ∫ α ( s ) e − r s dt
0
t
I 2 (t ) = σ ∫ e − r s dW ( s ) ,
0

G
GRC5MP, March 2006
Seite 33 12345
What may we hope for?
A simple model for strategic risk
t
Properties of the Ito Integral I 2 (t ) = σ ∫ e −r s
dW ( s ) :
0

• it is a martingale, in particular E[ I 2 (t )] = 0 .

• the variance is given by


t
σ 2
var[I 2 (t )] = E[ I 2 (t ) ] = σ ∫e
−2r s
2 2
ds = (1 − e − 2 r t ) ,
0
2r
• it is normally distributed.

G
GRC5MP, March 2006
Seite 34 12345
What may we hope for?
A simple model for strategic risk

Distributional properties of P (t ) = I1 (t ) + I 2 (:t )

For t > 0, the distribution function of P (t ) is a normal distribution with


t
E[ P (t )] = ∫ α ( s ) e − r s ds
0
σ2
var[P (t )] = var[ I 2 (t )] = (1 − e − 2 r t ),
2r
1 − exp(−2 r t )
VaRκ = Φ (κ ) var[P(t )] =
−1
σ Φ −1 (κ )
2r

EaRκ
Φ is the standard normal distribution function and κ the confidence level
G
GRC5MP, March 2006
Seite 35 12345
What may we hope for?
A simple model for strategic risk
Limits:
• For t → ∞ strategic VaR simplifies to
1
VaRκ = EaR(κ ),
2r
1 − exp(−2 r t )
• For small values of r we obtain 2r
3
r→0

VaRκ = ( )
t + O(r ) ⋅ EaR(κ ),
2.5

2
5

r = 0.1
1.5

0.5

0
0 2 4 6 8 10
time
G
GRC5MP, March 2006
Seite 36 12345
Thank you very much
for your attention!

G
GRC5MP, March 2006
Seite 37 12345

Potrebbero piacerti anche