MAS Finance meets
Bank Julius Baer
Presentation of
B. Hodler/N. MacCabe
April 2, 2004
Agenda
Julius Baer Group
Risk management organisation
Risk landscape
Working with a MAS Finance intern: a case
study
Questions / Discussion
Julius Baer Group (figures in Mio CHF)
2003
Assets under Mgt
Net operating income
Net profit
Equity
Capitalization
Headcount
ROE
115,500
1,020
82
1,474
4,282
1,766
5.3 %
1995
49,400
525
113
1,164
1,514
1,470
10 %
Julius Baer Group
Private Banking
Asset Management and Funds
Trading
Corporate Center
Risk Management
Finance and Controlling
Legal and Compliance
IT and Operations
Communication
Human Resources
Investment Research
Risk management organisation
Board of Directors committees:
Risk committee of the board (quarterly)
Audit committee of the board (quarterly)
Executive Board committees:
Group ALM committee (monthly)
Group risk committee (weekly)
Group lead management committee (on request)
Risk management organisation
Group Risk Management
B. Hodler, CRO
A. Weber, Deputy
Risk Advisory
N. MacCabe
Support
M. Calpini
Private Banking
D. Münchbach
Credit Risk
A. Weber
GRM NY
HR Würgler
Relationship
Mgt
K. Schmid
Market Risk
S. Altner
Operational
Risk
B. Hodler
Asset Mgt &
Funds
B. Briner
Trading
R. Winkler
IT & Operations
U. Läderach /
Ph. Malherbe
J. Hüsler
Julius Baer Group Risk Landscape
Strategic / Business Risk
Operational Risk
Market Risk
Funding / Liquidity Risk
Credit Risk
Clients & products
Execution, delivery & process
Fraud
Personnel
Legal & tax liability / default
System & physical risk
Reputational Risk
Six commandments of risk management
Foster risk and return awareness
 Understand your profits
 Be prepared to pay
 Reconcile with diligence (and on time)
 Track the cash
 Watch your systems

Case study
Finance practitioners and academia working together
Project to model issuer specific risk on non-government
bonds at Julius Baer

What is issuer specific risk?

Key advantages of approach taken

The practitioner’s perspective

The intern’s perspective
What is issuer specific risk?
Risk from changes in price of a bond NOT due to
changes in the risk-free rate of interest


Issuer-specific risk (ISR) present in all non-govt bonds
Comparable magnitude to pure interest rate risk – can
be much larger


Modelling pure IR risk fairly easy

Modelling ISR much harder
Problems with modelling ISR

Reliable historic prices are not available for most bonds

Even if they were available they would be of limited use because
time to maturity of a bond changes every day

Theoretically, problem 2 could be resolved by building a yield
curve (based on numerous bonds) for each issuer. Very difficult
in practice and very time consuming.

An approach based on the rating (S&P, Moody’s) of a bond could
be used, but this presents numerous difficulties too
How did Enrique model ISR?

Measured spread of each bond (at current market price) over risk
free rate at same time to maturity (TTM)

Captured not only risk free yield curve for each currency, but also
various rating specific yield curves per currency (from
Bloomberg)

Took the interpolated spread over the risk free yield curve at
each TTM and for each rating specific curve

At each TTM calculated the historic volatility of these various
rating specific yield curves

Used discriminant analysis to determine probability that each
bond’s spread would fall into a given rating category (usually
several probabilities, summing to one)
How did Enrique model ISR? (2)

Constructed an expected spread history for each bond (based on
historical spreads of each rating category and posterior
probabilities)

Once the expected spread history was calculated, GARCH was
used to find the best fit for the time series. These then drove
simulated paths for the expected spread history. This had effect
of rewarding diversification.

All of this was then automated in a routine using the SAS
statistical package
Key advantages of this approach

Rewards diversification

Backtesting against actual bonds (with reliable history) shows
model makes good estimates

No additional data on individual bonds needed

Can deal with any bond

Routine chooses best GARCH model for each bond‘s expected
spread history

Because main input is bond‘s current spread, model reacts
immediately to changes in market perception of an issue‘s credit
quality.
Financial practitioner‘s perspective

Assign one clearly defined task only to the intern

Task should require developing new approach to some problem
(e.g. a modelling problem)

If modelling involved, define an approach to backtesting early on

Recognise you are taking a risk

Encourage intern to attempt multiple approaches (unlikely to be
right first time)

Review progress regularly (at least once a week)

Be prepared to spend time helping the intern

Ensure intern has time to write thesis.
Intern‘s perspective

Ensure task is clearly defined and that you understand it

Ask yourself seriously if you have what it takes to do the job

Try to gauge whether the task is doable in the time

Find out who your supervisor will be and make sure you spend
time talking to them about project. Can you work with them?

Ask how much time your supervisor will be able to spend with
you.

Ensure you have time for writing your thesis
Expected Spead History Calculation
Position‘s Rating may change during its lifetime. Thus, given position‘s
current YTM, a Discriminant Analysis was performed using the
simulated changes
Probabilities of „membership“ into each Rating Category are obtained
and these are used to construct an Expected Spread History (ESH) as
follows:
ESHt = IssuerSpread + Current RFR*ExpChanget
C
 PIssuerSpread  Category * ChgCat
ExpChanget =
i 1
Bond:
TTM:
IS:
3.75 Akademiska 06
2.09 years
19.17 bp
Group
Probability
AAA
0.0000
AA
0.6224
A
0.3776
i
i, t
Monte Carlo Simulation and Risk Measures Calculation
Using Monte Carlo, two bonds with exactly the same TTM and YTM will
have different simulated spreads. In this way, the ESH of this simulated
paths will not be perfectly correlated and diversification reward is
attained.
For each trading day, a random number from a (0, st2 ) is drawn. The
simulated pahts consider the volatility‘s time dependence.
Changes in the PV of the position is calculated using the Simulated
Spread.
Backtesting
Some bonds issue in CHF were selected with its price past history, and
a daily HSVaR was computed for the last 210 days.
Changes in bond‘s price due Issuer Spread is isolated and compared
with the HSVaRs.
HSVaR99
Exceptions Simulation
Rabobank
Hessen
General Motors
BP Amoco
Roche
Gemeenten
Electricite de
France
HSVaR95
ExpSpread
Simulation
ExpSpread
# days
Observed
Expected
%
Observed
Expected
%
Observed
Expected
%
Observed
Expected
%
Observed
Expected
%
Observed
Expected
%
1
2
0.50%
1
2
0.50%
1
2
0.50%
1
2
0.50%
0
2
0.00%
0
2
0.00%
2
2
1.00%
2
2
1.00%
5
2
2.40%
2
2
1.00%
1
2
0.50%
0
2
0.00%
5
10
2.40%
3
10
1.40%
6
10
2.90%
3
10
1.40%
2
10
1.00%
0
8
0.00%
12
10
5.70%
8
10
3.80%
19
10
9.00%
10
10
4.80%
5
10
2.40%
1
8
0.60%
210
Observed
Expected
%
6
2
2.90%
7
2
3.30%
13
10
6.20%
24
10
11.40%
210
210
210
210
210
170