Scheffmann, Tim, "Bank Management of Risk Dynamics", 2005 July 17-2005 July 21

Bank Management of Risk Dynamics
Tim Scheffmann
IBM -— Business Consulting Service (BCS)
An den Krautgaerten 29, 65760 Eschborn, Germany
Mobile: +49 (172) 7929242
Tim @Scheffmann.de

In banks decisions are made in a speedy and complex environment often with huge un-
certainty. This risk must be managed proactively on an enterprise level. Thereby a sys-
temic view of the bank is essential. Up to now there is no standardised approach for
analysing overall risk dynamics of a bank. Most risk models are constrained by their
static view, so that they hardly capture the rapid and discontinuous changes. This pa-
per examines the dynamics by applying system dynamics to enterprisk risk manage-
ment, with the aim of understanding the banks’ risk dynamics. In order to simulate the
risk dynamics an enterprise risk model was developed. By combining the disciplines of
enterprise risk management and system dynamics, this paper shows how a systemic
view can improve structures in bank risk management.

Keywords: Enterprise Risk Management, Dynamic Modelling, Bank Management

“Risk is a choice, not a fate”
Prakash Shimpi (President Fraime LLC)

1. Introduction to Dynamic Risk Management

1.1. The Need for Dynamic Risk Management

Today’s bank managers make decisions based on incomplete information, price in volatile
markets and use information of complex systems. To accomplish this, risk management helps
to deal with uncertainty in the best possible way. The goal is to strike the balance between risk
and reward.' Financial risk management mainly considers earnings volatility in a bank’s credit
and trading department. Yet the word risk has two meanings.” First it is defined as an exposure
to a chance of loss or damage, which is a quite negative connotation. In the same instance, risk
is a function of likelihood and consequence. It denotes a revealing positive outlook: to take a
risk in expectation of a favourable outcome.’ We can use our marvellous capabilities to reason
on past events and convert the unknown future into an opportunity. In banking the term “risk”
is generally used to describe the likelihood of a loss or that an investment return is being lower
than expected. However, especially investment banks also make profits by actively taking risks
from other counterparties. After all it is not possible to make any gain without taking a risk.
Therefore, risk is on equal footing with opportunity and threat.

Each bank has a business specific risk profile and executes decisions with a different risk appe-
tite. The trading division of an investment bank close to Wall Street has a greater risk appetite

1 Leslie Rahl: “Lessons learned”, Risk Insights, 01/2002, available at www.cmra.com

? Definition of risk available at http://www.cogsci.princeton.edu/cgi-bin/webwn2.0?stage=1 &word=risk
accessed 28 January 2005.

3 The word risk derives actually from “risicare”, which in Italian means to dare or to have sufficient cour-
age. It originally meant “to circumnavigate a cliff”. In this sense risk is a choice rather than a fate. See also
Peter Bernstein: Against the Gods — The Remarkable Story of Risk, 1998, Wiley, p. 8.

than a small retail bank in Oslo, for example. Major Banks easily absorb huge risks, either be-
cause they have a strong capital basis, or because they possess the expertise to hedge risks fol-
lowing the Markowitz diversification concept.’ Regardless of their size, banks have to focus on
risk management at an enterprise level. In the past, risk managers oversaw hidden hazard and
liability risks; internal auditors dealt with accounting issues; business units focused on project
risks; and treasury handled foreign exchange and interest rate risks.” Although today these func-
tions are united, still risk silos are maintained by single units risk measurements. Risks can be
assessed at three levels: at the transaction, business unit and corporate level. ERM shifts the
focus of risk management towards a company-wide focus with defined responsibilities. While
traditional risk management works best for financial risks (i.e. transferable risks), ERM, by
contrast, stresses the management of operational and strategic risks. By the enterprise view we
are able to manage a risk portfolio with the tendency to optimise the overall risk. While some
risks hedge others, it is essential to capture the correlations between them. Only if all risks are
properly integrated can single strategic decisions can be made with a chosen risk appetite. For a
CEO to be able to manage the overall risk of a bank, a framework is needed to integrate single
risks and business lines.

This paper focuses on enterprise-wide risk management (ERM) as “an integrated framework
for managing risk and risk transfer of one bank in order to maximize firm value.”° Therefore,
the system boundaries are set like it is shown in the yellow part of figure 1. The reader has to
differentiate between the closely =

related term “systemic risk” and
the term “risk dynamics” which
denotes the stability of a whole
banking industry in a geographical
region.’ The main focus of this
study is on the risk dynamics of
an individual bank. Nonetheless,
at some points we will also con-
sider systemic risk for the whole
banking industry, since this also
has an impact on the risk of each
individual bank.

Figure 1: System Boundaries and Sub-systems

The driving force behind ERM initiatives are the regulatory requirements. As Robert Levine
notes, “A combination of regulatory and commercial pressures is driving organizations to spend
more than ever on technology to manage risks.”* Enterprise risk management is clearly driven
by evolving regulatory requirements such as the modified Basel Capital Accord or the Sar-
banes-Oxley Act (SOA). Section 404 of the SOA requires management to attest to the sound-
ness of a company’s internal financial reporting. The Basel Committee on Banking Supervision
of the Bank for International Settlements (BIS) has introduced more sophisticated requirements
for credit and operational risk management in the new accord.’ This new accord, known as
Basel 2, calls for a risk management system with risk sensitivity capabilities and in which eco-

4 Heinz Hockmann and Friedrich ThieBen: Investment Banking, 2002, Schafer Poeschel, pp. 74 et seq.

> Russ Banham: “The Art of Measuring Risk”, Reactions, 11/2003, Vol. 23, Issue 11, p. 55.

6 James Lam: Enterprise Risk Management — From Incentives to Controls, Wiley, 2003.

7 See working papers at the World Bank: http://econ.worldbank.org/view.php?id=39805, accessed 4
March 2005.

8 Robert Levine: “Risk Management Systems: Understanding the Need”, Spring 2004, in Information Sys-
tems Management, p. 31.

° The Basel II framework can be accessed under http://www.bis.org/publ/bcbsca.htm.

nomic and regulatory capital are aligned. However, as Figure 2 shows, ERM in banking is not
only driven by legal requirements, but also internal forces such as technological advances or the
finance industry itself.

Market place

Volatility, Changing Parameters
New Products (e.g. Derivatives)
Utility and Portfolio Theory
Rating Agencies

Enterprise-wide
Risk Management
in Banking

Finance Industry

Globalisation/Integration
Intensifying Competition
Focus on Value Creation
Customer Management

Legal Framework

IFRS

Basel II
KWG/ SEC Regulations
Sarbanes-Oxley Act

Technological Advances
Computer Trading
Process Documentation Tools
Service-oriented Architectures

Figure 2: Drivers of Enterprise Risk Management in Banks

ERM improves three essential, closely interlinked parts in banking: value creation, capital
structure and capital budgeting. To create value, investments need to be budgeted. If value is
created, the capital shifts due to the generated profit. The upcoming revenue stream allows the
bank to budget new projects with the aim of creating new value with a certain risk. “Managing
risk at the corporate level can increase the value of a bank, because it can reduce [...] costs of
equity and debt as well as that of transaction costs.”'° Bank managers are able to reduce the
cost of capital by better risk management approaches. In order to build such an approach, the
bank must select effective strategies of risk elimination, risk transfer and risk taking. Figure 3
lists some of the more recently used risk handling techniques:

x Instruments
nissionpracel Hedge/Sell__|Diversify Insure Set Policy Hold Capital
Eliminate/Avoid Xx x Xx Xx
‘Transfer Xx Xx
Absorb / Manage x x x

Figure 3: Common Risk Approaches and Instruments in the Finance Industry"!

The dynamic perspective of ERM offers an enormous opportunity not only to make risk optimi-
sation results visible, but also to accomplish sustainable business achievements. ERM is be-
coming ever more important owing to the increasing complexity of risks. Globalisation, for
example, has added new risk components such as country risk to a business. Risk components
are also changing, making a risk overview necessary. A bank’s risk portfolio is similar to its
investment portfolio. The portfolio view of risks allows the bank to make strategic decisions as
to which risks should be eliminated and which new ones can be taken on board. Today’s bank-
ing world is swiftly changing due to innovations in instruments and products. Therefore, ERM
has to transform risk management from a rather static activity to a dynamic, proficient process.
The Greek root of the word dynamic, dunamikos, indeed indicates a powerful or energetic
change, marked by continuous or productive activity. The ultimate goal of ERM is to maximise

10 Gerhard Schrock and Manfred Steiner: “Risk Management and Value Creation in Banks", Risk Man-
agement, 1st edition, Springer, 2004, pp. 55 et seq..
11 Partly based on Schréck and Steiner, op. cit., p. 56.
the firm’s value by continually optimising risk consumption for a certain return. By improving
the dynamic capabilities of enterprise risk systems, a bank can gain valuable insight into the
correct level of risk exposure on a flexible and permanent basis.

In the world of finance, models are dynamic in a different way than in the system dynamics
view. Most financial models are considered dynamic, and they are indeed designed to contain a
certain degree of change over time. However, the structure of the modelled revenue stream is
often not changed at all, and financial models remain to a certain extent static, meaning that the
full impact of forces is not captured, even though the behaviour changes over time but within
the same structure. Systems and especially dynamic models can vary in terms of the degree of
dynamism.’? Dynamic systems behave in a certain way, which means either that they do some-
thing, or that interaction is taking place within themselfes. These dynamics are internal and
external. As far as the author is aware, this thesi the first to examine the dynamics in finan-
cial risk management from a systemic viewpoint. Taking a systemic view allows to capture
complexity, which is hard to model in existing risk models. This paper extends research in two
ways:

1. It provides insight into the dynamics of ERM. It is not enough to calculate static risk
indicators, since risk changes drastically over time. Banks need to implement these risk
dynamics into their models in order to price risk exposure correctly.

2. Strategies, patterns and systems features are examined with regard to how a bank can
reduce the overall risk (dynamics). The processes that have to be implemented to ab-
sorb risk are also analysed.

A system dynamics model is developed to find answers to questions like: What kind of risk
dynamics does a CEO have to face in an ERM framework? What kind of dynamic behaviour is
strongest, and what impact does it have? What mostly makes the bank’s economic capital in-
crease? By applying system dynamics methods to ERM, we are able to capture risk dynamics.

The purpose of the model is first to discover the dynamics in an exemplary bank within an en-
terprise risk framework, and then to derive ways to improve the risk system.

1.2. Risk Structures

The most important risks a bank faces are credit, market and operational risk. Alongside these
are risk types such as liquidity risk, funding risk or country risk. Figure 4 shows the most im-
portant risk categories.

tow risk C) - : :
High risk @ A Universal Bank’s Risk Profile
@| © © K) 0 |
Credit | Market Operational [ Business | Other
Counterparty Trading loss / Damage Margins, Liquidity,
default price uncertainty | ___ Technology Volume Reputation
Personnel
| Organisation

External

Figure 4: A Universal Bank’s Risk Profile

12 Jiirgen Strohhecker: “The Object-oriented Model of a System's Dynamic”, in “System- und objektorien-
tierte Simulation betriebswirtschaftlicher Entscheidungen”, Issue 53 Scientific papers of the Industriesemi-
nar University Mannheim, Duncker & Humblot, Berlin, 1998.
Risk as an exposure to a chance of loss can be measured in various ways. A loss event occurs
when asset values change to the negative. Since Markowitz, the risk of an investment has been
measured by the standard deviation of outcomes.’ In doing so, probabilities of loss are deter-
mined from historical data. The probability determines the degree of certainty that can be as-
signed to certain events (e.g. credit defaults) happening within a specified time interval or
within a sequence of events. The most common industry standard is the VaR concept. The
value at risk is the value that will not be exceeded (e.g. in the next 10 days) with a probability
of 99% (the confidence level= 1-«). For a given portfolio, VaR measures the possible future
loss which will not be exceeded with a high probability for a certain period:

VaR =(o" (l-@): svanne J Havin J VALU, irony » Where @'= the inverse standard normal

cumulative density function, o= the standard deviation and t= the mean (expected return), for 1
day as a holding period.'*

1.

Credit Risk: Pricing Credit Defaults

Credit risk is the oldest and, in terms of economic capital, the most important risk for banks. It
arises from defaults, namely when a debtor does not fulfil a contractual payment obligation to
the bank. Often credit risk is the largest single risk a bank has to face. Credit risk is sudden and
its potential impact is huge.'* It is now more important than ever to assess the credit quality of a
client (issuer or borrower), an assessment known as rating. Between the actual counterparty risk
and the perceived risk is a time delay, as it takes time for the bank to recognise the actual de-
fault situation of the customer. At the time of the default, the bank recognises the exposure
amount, which can be reduced by selling the collateral (recovery rate). If the exposure is not
covered fully, the bank suffers the loss rate of the credit exposure amount.

Economic
Situation

{ ____-» Counterparty Credit

<3 aa * Risk *|_ Risk

Default a
Probability Ve ty

Credit Recovery
Exposure Rate

Figure 5: Credit Risk Components'®

Interest Rate Change
- Fixed Income

1. Market Risk: The Known Quant World *%
Market risk arises from the possibility of losses Coe fie F x =
resulting from uncertainty about changes in el et
market movements (e.g. equity prices, interest Equity Price— + he
rates, exchange rates, commodity prices, etc.). Changes
Most of a bank’s market risk resides in the trad- Foreign Exchange EX) -
ing group. Change in Currency Trading

B EoUe Relations Volume

Figure 6: Market Risk Driver

18 Harry M. Markowitz: “Portfolio Selection”, Journal of Finance, 1952, No 7, pp. 77-91.

14 More information about the VaR concept can for example be found in Kevin Dowd: Beyond Value at
Risk, Wiley, 1998. This paper gives only an overview of risk measurement methods.

15 Richard K. Skora: “Modern Portfolio Credit Risk Modeling”, available at
http://www.skora.com/modern.pdf, accessed 6 March 2005

16 The two parallel lines “||” in the figure indicate the time delay.
Depending on the transactions, there is more risk either in the trading book or in the bank’s
book.'’ Market risk emerges from the mismatch of demand and supply, and is also influenced
by the correlations of the changes and their levels of volatility. A bank can lose money in a few
seconds (e.g. buying currencies high and selling low), but losses also occur due to long-term
market trends such as collapsing real estate prices or changes in interest rate levels. Market risk
also occurs when, for example, investors are no longer willing to pay a high price for a certain
stock. To illustrate this, Figure 7 shows how the Nasdaq index lost 65% of its value between.
March 2000 and March 2001 as a result of the bursting of the internet stock bubble.

6000 5500

5000
5000

4000

3000

2000

1000

0

1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004

Figure 7: Long-time Chart and Radical Fall Chart from the NASDAQ”™
1.2.3. Operational Risk: Risk-Resistant Operations

The Basel Committee on Banking defines operational risk as “the risk of direct or indirect

losses resulting from inadequate or failed internal processes, people and systems or from exter-
nal events.”'? Operational risk is the threat to the bank’s operations. Unmanageable events
(such as September 11), insufficiently defined controls and project failures increase operational
risk. Business disruption, processing risk as well as fraud and mistakes are included in this risk
type, too. In 1997, for example, the National Westminster Bank incurred enormous losses ow-
ing to badly handled operational risk. The bank lost $127 million, and “had to greatly reduce its
trading operations, because its options traders had been using the wrong data for implied vola-
tility in their pricing models, and was therefore taking risks that they did not see.””? Operational
risks are always industry-specific. Banks handle money, and their business is based on trust and
competence. Making mistakes certainly puts off new customers. Financial institutions have to
operate accurately, must implement dual controls, must be proactive in combating fraud, and
must double-check transactions.

System

+ Failure

ee i

External Event . Human +| Operational
Causing Loss + — Error "| Risk

wt

Fraud~
Process —~
—— Alignment

Figure 8: Driver for Operational Risk

17 According to Uwe Schulz, Value & Risk AG, mortgage banks tend to keep speculative interest positions
in the bank's book, so that no regulatory capital has to be allocated.

18 Data source: Yahoo Finance:

http://finance.yahoo.com/q/hp?s=%5EIXIC&a=00&b=1 &c=1984&d=01 &e=258f=2005&g=d, accessed 25
February 2005.

19 BIS, available at http://www.bis.org/publ/bcbsca.htm

20 Chris Marrisonop. cit., p. 6.

A challenge for calculating operational risk is the lack of data. Operational risk is typically
caused by a variety of factors, including natural events, human errors and technical problems,
and the outcomes of these events are often unknown. Some mistakes go completely undetected
because they are small enough to hide, but embarrassing enough to admit to. For operational
risk, therefore, a more complex and mixed VaR approach is used, because the loss data avail-
able are insufficient to apply the normal VaR concept.”!

1.2.4. Business Risk: The Set-up

Business risk is regarded as uncertainty if the bank’s (expected) revenues are able to cover its
expected fixed cost base and variable expenses. Business risk represents the uncertainty of
earnings due to changes in business conditions, namely market environment, client behaviour
and technological progress (volume and margins). Risk arises at a time when revenues are for
example declining faster than adjustments can be made to the cost base. Business risk reflects
the bank managers’ decisions. Products, sales and prices have to be managed well to guard
against business risk. The three key drivers of Interest Rate

business risk are as follows: Spread

1. The volatility of revenues due to inten-
sifying competition, economic cycles

and their effects on the customer base Business
or the lifecycle of the bank’s products. Satisfied prank isk
2. Business risk tends to grow in propor- Customers er aume
tion to a bank’s fixed cost base. as
3. The volatility caused by the variable $s Force
costs. pen

Figure 9: Business Risk Drivers

Business is about people. In a competing environment, financial institutions try to ensure high
business volume by pushing and training their sales forces. If a customer is wrongly advised on
financial products, however, this will lead to shrinking business volume in the future, since the
customer loses confidence and the bank suffers an image loss. It benefits the bank very little if
s sales force is ingenious, but they sell the wrong products. Closely connected to Business

Risk is also reputation risk. Banks sell liquidity and asset availability to clients. Accurate finan-
cial transactions and discretion are a must for the banking business. The minute these business
attitudes are called into question, clients loose their confidence, and banks their customers.
Reputational risk has an instantaneous impact, as indeed can be observed during a bank run.
The concept of Reputation Risk is known in the System Dynamics approach as Word of mouth.

Research conducted by the Royal Bank of Canada found that the stock market tolerates one-off
mistakes. Typically, when bad news is announced, the bank’s share price drops but recovers
within about 90 days. However, there was no tolerance of strategic mistakes or contradictions
to a company’s formerly chosen risk appetite. It took a year for a bank to re-establish its credi-
bility and to restore its share price.” Investors’ confidence is also shaken by deficiencies in
corporate governance, which can lead to costly fines and legal bills as well as significant dam-
age to the institution’s reputation, with the end result invariably reflected in the share price
taking a severe pounding. Establishing and maintaining a good reputation requires foresight,
but is definitely worth the effort. Therefore, in recent years banks have established communica-
tion departments that are responsible for presenting the bank positively in the media.

21 Agatha Kalhoff and Marcus Haas: “Operational Risk - Management Based on the Current Loss Data
Situation”, Operational Risk Modelling and Analysis, 2004, pp. 10 et.seq..
22 Russ Banham: “The Art of Measuring Risk”, pp. 55 et seq..
1.3. Risk Aggregation and Integration

The central figure in managing risk is that of the bank’s economic capital. This states the
amount of equity capital needed to absorb unexpected losses with a high degree of certainty at
any given date. The certainty is chosen by each bank individually (normally > 99.95%). Eco-
nomic capital consists of aggregated risks and mostly takes into account diversification effects.
It does not equal the regulatory capital (the minimum which is required by regulation). In con-
trast, economic capital is the capital that shareholders would choose in the absence of regula-
tion, i.e. the amount that is needed to keep the bank operating during the next year and to main-
tain a certain rating.” It focuses on its buffer function against future unidentified losses. In gen-
eral, capital is the means to achieving the optimal capital structure, and also provides insurance
for the bank’s safety.

All stakeholders are collectively interested in not overstepping the critical threshold, where the
bank has to discontinue operations (i.e. a bank run). Figure 24 below displays the insurance
function of economic capital. The peak of the distribution with the highest probability is

where Gains > 0, as the bank should be expected to generate profits. In the area where there
are no gains, only losses, shareholders lose money because they own the company and the as-
sets in it. The different risk types of credit, market, business and operational risk are aggregated
to calculate the overall loss distribution as shown in the next picture.

Market Risk

Praha at

Probability of loss
greater than W,,

Credit Risk

= Operational Risk
Loss 0 Prosi Psbilty
Eyed aaa
ees — Low freque as
_— Wilh high impact come~
poe eeee ri eee vi -
xgeoet Law (El) Homi os"—sl Overall Risk eo —
=nieeey onan, ‘Value at Risk atthe — =
eis
Critical
Threshold
for 1k Run
ees
Le lee
LEB:
(ee Credit Risk
eligi
8 3:
Le 3 Shareholders
Losses 100%

Economic Capital

Figure 10: Economic Capital as an Insurance Function”

23 http://www.defaultrisk.com/pp_super_41.htm, accessed 27 February 2005.
24 Note: The expected return should be farther to the left than actually depicted. Figure based on Schroeck
and Steiner: op. cit., p. 158.

Since VaR measures risk as a negative deviation from the expected outcome, it is a good meas-
ure for the bank’s total risk and can be used as a common “currency” to quantify each risk type.
Economic capital is therefore a function of the riskiness of the bank’s activities and the bank’s

desired likelihood of solvency: Pix - E(X))> economic capital |< a.

This depicts the probability that the distance between the expected outcome X and unexpected
deviations E(X) will not exceed the economic capital. This guarantees a certain solvency, with
an . until year-end.** The overall bank VaR can only be derived indirectly from the different
divisions in a bank, since they calculate risk parameters differently. For example, the VaR for
market risk is typically calculated for a ten-day holding period”®, while VaR for economic capi-
tal is based on a 250 trading day basis. Catastrophic risk with large losses occurring only with a
small probability violate the coherence of VaR. VaR is not sub-additive, i.e.

VaR(X + Y) # VaR(X)+VaR(Y). As Andreas Krause points out: “In many cases the condi-

tions for subadditivity are (at least approximately) fulfilled and VaR is nearly coherent. But
remarkably, the payoff distributions of options and similar derivatives mostly violate these
conditions. And it was the use of such instruments that caused many of the large losses in the
early 1990s and led to the development of VaR.”””

The economic capital over the single risk types is in practical terms aggregated by adding the
capital requirements to determine the overall amount. In actual fact, this practical approach
ignores two correlation matrices: the correlation within a risk type, and the correlation matrix
across risk types.”* Given that risk types influence each other, these effects should be consid-
ered. Nonetheless, one could argue that even if we neglect the correlation effects, the results are
still valid, as shortly before a bank’s insolvency, the correlation parameters change drastically.
Banks determine correlations for the second matrix by employing macroeconomic simulation
models which are also used to enhance the operational risk model. Such models are complex
and modelled only with difficulty. Large events are so rare that risk relations cannot be deter-
mined on an ad hoc basis. A few more events can quickly invalidate any a priori modelling
decision. Risk systems in extreme situations show a very different behaviour than during nor-
mal events. A natural catastrophe or terrorist act can wipe out a stock market, whereas there is
no correlation among the single shares. “Large risks are much more likely to be interconnected
because the large-scale processes they unleash will overlap.”

There is also a systemic risk in the VaR measurement, which could lead to a self-fulfilling
prophecy of doom. Actions taken as the result of a VaR estimate, e.g. an asset liquidation,
could themselves precipitate a crisis. In this instance, VaR becomes obsolete as a risk measure
because the distribution of outcomes changes significantly. This effect is not taken into account
in the VaR estimation. With many market participants acting in the same way, namely trying to
sell the asset, VaR may lead to loss realisation and trigger a domino effect.*® Figure 12 shows
the economic capital development at Deutsche Bank. Economic capital is measured from 1998
onwards, first without any information about the components. Between 1998 and 2002 it stead-
ily increased. The drop in 2003 was explained in the annual report by improved credit portfolio
quality, better market prospects and the ability to react on the cost side in weak market periods.

25 Ibid., p. 169.

26 The regulatory process requires a confidence level of u=99% and a ten-day holding period. Internal
models also use a one-day holding period and a confidence level of a=95% for back-testing reasons, and
also because the economic P&L is calculated daily.

27 Andreas Krause: “Exploring the Limitations of Value at Risk: How Good Is It in Practice?”, Journal of
Risk Finance, Winter 2003, pp. 19 et seq..

28 Schroeck and Steiner: op. cit., p. 210.

29 Jim Brander and Sam Manoff: “ERM and DFA Using Active Knowledge Structures”, paper available at
http://www.casact.org/pubs/forum/03sforum/03sf001.pdf; accessed 15 January 2005.

30 Andreas Krause: op. cit., pp. 26 et seq..

25

2089 22-43

Economic
Capital (bn. €)
pital(bn.€) wy | _________---_ gy ----- eBusiness
Risk 79%
Credit Risk mone
Risk 14%
15
Market Risk
(incl. Equity) acre pee
Operational Risk 1°
Business Risk 5 ating

35%

Balance

1998 1999 2000 2001 2002 2003

Figure 11: Economic capital development at Deutsche Bank 1998 — 2003

On the next page in figure 13 is the economic capital for 2003 and the parameters for the four
leading private banks in Germany pictured. The results are not scaled to an overall confidence
level & to show the original data. To make the figures comparable, diversification effects have
to be estimated where the necessary information is not provided in the annual risk reports.

Comment on Figure

Period a Correlation/Diversification B
Will be implemented from _| Diversification effect
: “ of
Deutsche Banc Lyear (29.98% 2004 onwards. estimated at 22%.*!

Commerzbank lyear |99.95% | Effects considered.

Dresdner Bank lyear | 99.93% | Effects considered. i

Annual Risk Report values | Values for single risk
Hypovereinsbank | 1 year | 99.95% | are published after divers components scaled up

fication/portfolio effects. by estimation (22%).
_ 20
& 8p
Credit Risk = et-
= yat-
@ Market Risk eB
10
8
1 Operational Risk 5
4
Business Risk ar-
0 Cue
24--| 367 219 |
Diversification Effect
+ Deutsche Bank |Commerzbank [Dresdner Bank IHypovereinsbank
Economic Capital 31.12.2003 (mn. €) 16,674 10,100 10,600 12,128
lm Credit Risk 7,968 4,300 4,800 3,316
[a Market Risk 5,912 4,400 3,700 5,956
| Operational Risk 2,282 900 1,400 1,664
IC Business Risk. AAN7 500, 700 1,192
[D Diversification Effect -3,668 -2,200 -2,400 -2,187

Figure 12: Economic Capital Components at the Four Main German Banks”

31 22% is the average diversification effect of the other two banks. While bearing in mind that this estima-
tion only serves as a first approximation, it nevertheless serves as a good benchmark (see also Figure 26
and explanation).
This chapter has considered how to derive economic capital based on the VaR concept. This
concept, although not without some minor weaknesses and therefore risks in itself, provides a
good approach for measuring the deviation from the expected value. Risk aggregation and cor-
relation have also been considered. To enhance the approach, more dynamic elements such as
business cycles, macroeconomic events and investor behaviour are to be added to the model.
The next chapter accordingly examines the dynamics inherent in banking risk.

"Anyone who believes exponential growth can go on forever
in a finite world is either a madman or an economist"
Kenneth Boulding

2. Risk Dynamics: From Equilibrium to Non-linearity

The real impact of risk is often separated by time and space from the event instigating it, so that
arisk manager has a hard time quantifying risks appropriately. Risks can develop either quickly
or slowly - and can sometimes gang up and have a cumulative effect. Risks may hedge, aggre-
gate with, magnify or be uncorrelated with other risks. Risk concentration is one of the most
dangerous types of risk, and can cause major insolvencies.

Real estate risks, for instance, tend to mount up over time, e.g. by uniting market, credit, liquid-
ity and business risks. Real estate risks have made many banks insolvent, as sub-section “Bub-
ble Dynamics” demonstrates. Such risks are highly contagious in two ways: first, in case an-
other bank takes over, the assets of the healthy bank will be corrupted; and second, the whole
economy can be hit if a real estate bubble bursts, affecting all market participants adversely.
One really large risk, or indeed a plethora of small risks, can put a bank out of business.” Risk
is by definition determined by its severity and frequency and therefore risks have different dy-
namics. Their severity varies from high impacts (e.g. a wrong strategic decision or all servers
down) to minor mistakes (like one incorrect number in a transaction). The frequency ranges
from “per second” to “once in a lifetime’. In risk management systems, behaviour such as “‘ex-
ponential growth” and “overshoot and collapse” can be found more often and for a longer pe-
riod than for instance goal-seeking behaviour, because prices move constantly. These dynam-
ics, as displayed in Figure 14, are considered in the following sub-sections.

System's Behaviour Appearance in Risk Management
“| Exponential growth Market growth, Overreaction, Reputation, Rating, IT
Goal-seeking Equilibrium, Laying off until employees are efficient
‘| Oscillation GDP cycle, Market cycles
S-shaped growth New product growth period, Efficient capital allocation
-| Overshoot and collapse _| Bank run, Bubble dynamics, Cover up mistakes, IT

Figure 13: System Dynamics in Risk Management
2.1. Unstable Equilibria

A system is said to be in equilibrium when there is no change on the macroscopic level and no
net forces on the system.™ Although there is no perceived change and no process of becoming
different, forces can nevertheless still be at work. The change may be too incremental to be
perceived, or a powerful negative feedback keeps the state of the system nearly constant despite

82 Data were derived from the 2003 annual reports of the individual institutions.

33 Mark Carey: “Changed World, New Risks”, presentation in 2003, slide 18, available at:
http)/www.delcreo.com/delcreo/free/docs/Changed%20W orld%2C%20New%20Risks.ppt, accessed 7
January 2005.

34 http://www.sasked.gov.sk.ca/docs/chemistry/ksc_b.html, accessed 3 February 2005.

countervailing forces.** Efficient market theory argues that all past information about a stock is
already inherent in the current market price, and that it does not help to evaluate future per-
formance. Nevertheless, the equilibrium price moves, sometimes without any news or informa-
tion about the stock, simply because an investor decides to buy a stock at a certain price. In
contrast to the efficient market theory, findings from behavioural finance show that charts of
the last year’s performance still influence actual investor’s behaviour.

Charts of stocks with a salient high in the past create expectations that the stock has the poten-
tial to perform better in the future. Investors are also more likely to keep these stocks than those
with a significant low in their price history. “Past prices exert a stronger influence on investing
decisions than trends.”*° This is an overreaction from the efficient market theory point of view.
Investors overreact to past information. They face an enormously complex situation when they
try to decide where to invest their money. One way to reduce such information complexity is by
using charts to compare investments. However, such charts can be a self-fulfilling prophecy for
the investor, although they are only based on historical information. By relying on past infor-
mation, investors become overconfident about future performance. Whether investors behave
rationally in pricing stocks or whether they overreact, prices are constantly changing, driven by
information and investor decisions.

Figure 31 depicts a system dynamics <Price> a Sacsved]- —
+ ’ ioe =~ vice Tn
model of a momentum investor’s de- Change in
cision-making. Momentum implies Bereaved Face. At
the partial incorporation of informa- — < Initial Price > Trend f Price
rey : : Time to perceive fl \
tion into stock prices. Such investors Price ( )

only care about the price trend in mak-
ing their decisions. In behavioural
finance research, informed investors

———$ Historical] — |
Change in Price

Historical Price

and herding investors are sometimes ~ Duration over which to
distinguished to describe the different calculate price trend
behaviours.

Figure 14: A Momentum Investor’s Decision-making”

2.2. Exponential Growth and Decay

Unstable equilibria are caused by a force in the market Fear of Not Being Able

pushing toward a higher or lower price. This behaviour to Withdraw Money

can be shown by a feedback loop. A feedback loop is +

an element of a system which (in)directly influences 4) =

itself. Two or more system portions are involved in the Bank's Withdrawals
loop, one with a correlating effect on the other. A Failure Bank Panic

feedback loop with an overall positive correlation

leads to reinforcing behaviour, e.g. overreaction and
unstable equilibria. Exponential growth patterns can be
detected throughout the whole bank system. One fa-

mous loop exhibiting such behaviour is that of a bank panic.

Bank's 4
Capital

Figure 15: Bank Panic Feedback Loop*®

35 John D. Sterman: “Business Dynamics”, 2000, McGraw-Hill, p. 127.

36 Thomas Mussweiler and Karl Schneller [both psychology professors at the university in Wurzburg]:
“What Goes Up Must Come Down — How Charts Influence Decisions to Buy and Sell Stocks”, Journal of
Behavioral Finance, 2003, Vol. 4, No 3, pp. 121-30.

37 Model developed by Getmansky and Papastaikoudi: op. cit.
The fear of not being able to withdraw money leads to panic, in which everyone tries to with-
draw their investments. By doing so, the capital basis of the bank is weakened until finally the
bank fails. Associated with this loop is reputational and business risk, namely the risk of losing
customers. The reinforcing power behind the loop leads to an exponential loss in the number of
customers.

The rating cycle creates a very strong feedback loop is, because two positive feedback loops
combine forces. A rating can become a self-fulfilling prophecy, as Figure 35 shows. Business
partners - like customers or trading partners — perceive the default probability of a bank by its
rating. Customers may become more confident based on a good rating, which results in im-
proved turnover. This effect may even increase over the next few years, as customers become
better informed about the reputation of a bank. Nonetheless, ratings have a much stronger im-
pact on the refinancing counterparties. The lenders to the bank keenly observe the rating, since
this is a clear indicator of the credit quality they are accepting. A bad rating clearly increases
the cost of capital since it is interpreted as representing a higher default risk for the counter-
party. Turnover and the cost of capital both influence profitability and subsequently liquidity,
which is observed again by the rating agencies.

The rating process determines whether a

bank faces defeat or will thrive. Further Rating from —_

research needs to be conducted into the we Agency * Default
dynamics in the rating process, to estab- Liquidity Rrobablity

lish whether a bearish economic period + 2 }

leads to worse ratings, and whether rat- pclae Customer's

ing cycles follow GDP cycles. So far, it Profitability Confidence

has only been observed that hard eco- AS munever— y
nomic times lead to higher default rates BN Bank's Refinancing
of debtors.” By assessing the bank’s a Spread
strength and robustness, rating agencies Cost of Shan

are extremely interested in establishing Capita —

whether a bank has implemented a risk
management system.

Figure 16: Powerful Loops Concerning the Rating of a Bank

Other examples of exponential growth concerning business and operational risk are cost explo-
sion and technological advancement. According to Moore’s law, computational power grows
exponentially. The number of transistors on a chip doubles every 24 months. Since a bank basi-
cally consists of information valued with money, or as John Reed, Citicorp., put it, "banks may
become nothing more than product lines of code in a big computer network”, software capabili-
ties are essential. If a bank is to compete in the market, as well as to ensure employee motiva-
tion, it has to provide fast data processing, and to manage costs. Mounting costs for software
maintenance in particular have become a major challenge for banks.

38 Adapted from the System Dynamics Group website: http:/www.systemdynamics.orq/DL-
IntroSysDyn/start.htm, accessed 11 March 2005.

39 John Frye: “Loss Given Default and Credit Portfolio Risk”, presentation at the Symposium on Enterprise
Wide Risk Management in Chicago, 26 April 2004, available at
http://www.casact.org/coneduc/erm/2004/handouts/fryebw.ppt, accessed 27 January 2005.

Transistors
[Mill]

2,000
1,800 + --------------------------
1,600
1,400
1,200 + -------------------------------
1,000

Year of
+: Introduction
1970 1975 1980 1985 1990 1995 2000 2005

Figure 17: Moore’s Law as an Exponential Factor for Bank IT”

Exponential growth patterns can also be found with regard to operational risk. Operational risk
managers try to minimise the number of mistakes made in operations, because errors increase
operational risk. As chaos theory has explained, small mistakes can have huge consequences, so
that costs increase exponentially — particularly when mistakes are not recognised for a long
period (e.g. calculating VaR with the wrong standard deviation for five years). Risk can also
increase through a series of mistakes, when these single mistakes are also interdependent. Mis-
takes can lead to even more mistakes if the pressure upon a person becomes too high. This
feedback loop, as Figure 37 shows, will only be tolerated for a short time, as costs quickly
mount. Costs

*
at

Pressure

D
Pressure

A ive No. of Mistakes

Workload
Figure 18: Working Pressure Feedback Loop and the Cost/Mistakes Relationship

Negative feedback loops are associated with a goal, so that the forces at work reach an equilib-
rium after some time. An example of this is the rate of employee fluctuation. A company lays
off a certain percentage of employees as part of cutting back, until it reaches the absolute mini-
mum number of employees needed to function.’

Figure 20 shows a negative feedback loop with Loan Quality
an associated goal. The desired loan quality
influences the default rate, since a bank cannot {
accept every default rate without becoming Loan Quality ‘er oval'v tah bai
bankrupt. The default rate influences the loan monn Celene
quality improvement programme, which hope- “4

fully will upgrade the loan quality until the ac-

tual loan quality and the desired loan quality will

reach the same level and an equilibrium is found.

Desired Loan
Quality

Figure 19: Loan Quality Loop with the Intended Goal

40 Intel Research Website, http://www. intel.com/research/silicon/mooreslaw.htm, accessed11 March 2005.
41 Helen Zhu: “Mental Simulation of Simple Negative Feedback”, p. 19, available at
http://sysdyn.clexchange.org/sdep/Roadmaps/RM3/D-4536-2.pdf, accessed 12 March 2005.

2.3. Bubble Dynamics: Overshoot and Collapse

A bubble is a self-fulfilling price escalation. Soap bubbles, which the idea of bubbles is based
on, usually last for only a few moments and burst on their own or on contact with another ob-
ject. They will always find the smallest surface area between points or edges. Thus, soap selec-
tively strengthens the weakest parts of the bubble and tends to prevent them from stretching
further, and also reduces evaporation to make the bubbles last longer. Thanks to their fragile
nature, bubbles have become a favourite metaphor for something that is attractive yet insub-
stantial.” This is very much the case with stock bubbles. These often appear around the turn of
a decade and typically exhibit the same psychological pattern: euphoria, greed, and hope at the
high point of the bubble, then fear and panic as the bubble bursts. Only after the bursting of the
bubble are investors ready for a sober assessment: before that they are caught in self-delusion
and blind to the mounting danger. In actual fact, we should be able to judge more accurately
based on our past and present experience. However, investors get carried away with excess cash
and buy stocks heedless of the consequences. The awareness that everyone is acting the same
way takes the form of a devotion to pure play, herding behaviour at its most obvious. Specula-
tion is unpredictable by its intrinsically chaotic nature.**

2000 New Economy
a a es es es I ee O ‘Speculation Bubble

(9 1995 Canadian Bre-x
Gold Bubble
1990 Japanese

~ 4 Stock Market Crash
1987 Stock Market
4° Crash NYSE

1980 Precious

Metal Bubble —

1929 Stock Market Crash

World Economic Crisis

ae ae San eae ae Sar ee (1920 Florida Real Estate Graze ~ - -

1822 London
O Loan Bubbie —

eae TulpMania
‘Amsterdam

1600 1650 1700 1750 1600 1850 1900 1950 2000 2050
Figure 20: History of Bubbles“

Figure 21 shows the main historical bubbles, where the size of the bubble is determined by the
geographical (local/continental/global) and monetary impacts. Historically the duration of
speculative bubbles has shortened, while the frequency has increased, perhaps owing to the
enhanced possibilities investors have nowadays due to the volume of information at hand. Most
bubbles appear at a distance of some years, which means that there are no shared memories of
past bubbles that could have arrested the momentum of a new bubble. The severity of price
bubbles and crashes in the economy is related to inexperience according to hypotheses in be-
havioural finance.** As time passes, new investors enter and old investors exit the market, re-

42 http://en.wikipedia.org/wiki/Soap bubble, accessed 24 February 2005.

43 Jérg Millenmeister, available at http://www.wallstreet-
online.de/ws/community/board/threadpages.php? &tid=0077 1 850&reverse=1 &page=8, accessed 23 Feb-
ruary 2005.

44 Datasource: http://www.caslon.com.au/boomprofile1.htm#railway, and
http:/www.investopedia.com/features/crashes/crashes1.asp, both accessed 4 March 2005.

4° Edward Renshaw: “The Crash of October 19 in Retrospect”, Market Chronicle, 22, 1988.

ducing the proportion of investors who remember the last stock market decline.“* The objects of
bubbles have changed in the course of time. Tulips, real estate, metals and stocks have all been
objects of speculation. These bubbles show different behaviours. The website of one investment
service company even classifies bubbles into four major types: dollar, money supply, stock
market and debt bubbles (see Figure 22 on the next page).

Fed Easy Money Policy

GSE's
Day Activities
Mutual Trading Stock
Fund Option
Trade
Purchases
Deficits L_Plans_|
Margin
(om (ise b/s
ul le
Defiation || Supply Stock
f Bubble Low Buybacks
a Savings Consumer /
TE Rate
fi Inflation ——
Consumer
internet Confidence
eCommerce
Wea Wall Street
Hype Hype

Figure 21: Bubble system”

Further research is needed at this point to establish how these different kind of bubbles interact.
Rather than focusing on macroeconomic interrelation, this thesis concentrates on micro-market
dynamics regarding risk in a single bank.

The behaviour of bubbles can be separated into three phases. First there is extreme exponential
growth, followed by a short period of equilibrium, and finally exponential decay. To describe
the growth of extreme bubbles, the volume formula for spheres can be used:**

4
V= 3 r?, where V is the volume and r the radius.

The volume of a sphere grows by the third exponent, if the radius is increased by only one unit.
The volume increases at enormous speed. Although bubbles have another kurtosis than globes,
this provides an indication of the speed.

After the escalation period, which is marked by high trading volume, a sudden break sets in,
resembling the lull before a storm. Then one person buys/sells at an unusually low price, and
the market becomes irritated.” These bids occur very suddenly and dramatically, which then
leads to a drop in the market. The price continues to rise, sell orders remain open and sellers

46 Gunduz Caginalp, David Porter and Vernon L. Smith: “Overreactions, Momentum, Liquidity, and Price
Bubbles in Laboratory and Field Asset Markets”, Journal of Psychology and Financial Markets, 2000, Vol.
1/ No 1, pp. 24-48.

47 http://www.cornerstoneri.com/newpage17.htm, accessed 1 March 2005.

48 Wolfram Research Inc., http://mathworld.wolfram.com/Sphere.html, accessed 23 February 2005.

49 An interview on trading experiments with Colin Camerer, Professor of Business Economics at the
California Institute of Technology, can be found in: “Market Efficiency of Bubbles”, Journal of Psychology
and Financial Markets 2002, Vol. 3, No 1, pp. 29-36 (32).

become nervous. The market follows the low seller. Panic sets in and then there is no longer
any stopping the price. The surface tension grows larger with a declining bubble growth rate.
Buy orders dry up near the peak until the bubble bursts. As Colin Camerer notes, “The largest
percentage changes are almost always drops rather than increases.”*° After the sudden burst, the
asset value drops crucially, sometimes even dragging along other asset classes. A decay pattern
mirrors the same behaviour as the growth observed first, but with a negative slope that resem-
bles a radioactive decay pattern. After the bubble has burst, rational behaviour resumes, at least
until the advent of the next bubble.

Figure 24 depicts the cause and loop structure. The two feedback loops influence the market
price. Speculation causes more and more money to be invested and creates more demand, while
some investors withdraw money because they are disappointed by the performance. Some in-
vestors try to counteract bad performance by the “when in trouble, double” loop, and buy even
though the asset price has failed to meet their expectations. However, the impact such investors
have on the demand is much smaller than a bearish market movement. The market volume can-
not be influenced by one transaction, unless it is significant enough.

Fresh Money
s—_—__
Investments
ay
Demand
Proft Speciiating § 7
Devotion
z cp |
Market Price a |
Expected Market x When in Trouble
wale: % a Double |
Awa Gey i
Asset Value Ma ‘Supply /
Gap Money +
withdrawal
Desinvestment

Figure 22: Speculation Market Dynamics

The bubble dynamics show classic

overshoot and collapse behaviour. The amount

ability to justify the exorbitant price is 2 2888888

eroded or consumed by the speculation

itself, which drives the price higher. = a A

This type of behaviour can be seen in oe '

past cultures which disappeared because 2 100005

they consumed the very resources they \

were living on. Line 2 in Figure 25 Y \ ,

represents the carrying capacity, while : opo| tt he Bice | :
line 1 represents the state of the system,  * att S00 To wa time

which feeds on the capacity.

Figure 23: Overshoot and Collapse Behaviour

5° Colin Camerer, ibid.
51 Lucia Breierova: “Generic Structures: Overshoot and Collapse”, 1997, p.10, available at
http://sysdyn.clexchange.org/sdep/Roadmaps/RM9/D-4480.pdf, accessed 12 March 2005.

A bank run illustrates just such an overshoot and collapse structure. It is a panic response that
occurs when a large number of people rush to take their assets out of a bank which they believe
to be financially unsound and about to collapse. This action, of course, usually causes the very
collapse feared in the first place, as no bank has enough reserves to cope with all its investors
simultaneously withdrawing their savings.” This vicious cycle came into effect during the
1930s, when an overall economic downturn caused the rate of bank failures to increase. “As
more banks failed, the public's fear of not being able to withdraw their own money increased.
This, in turn, prompted many to withdraw their savings from banks, which further reduced the
banking industry's capital reserves. This caused even more banks to fail.”**

An overshoot and collapse pattern can be detected with regard to credit risk when riskier bor-
rowers borrow more and more money until they become insolvent. This type of borrowing be-
haviour can be initialised by means of a rigid credit policy. Credit panic sets in, and credit lim-
its are exhausted. This cycle has caused some banks to be perceived to be lenders of last resort,
which inherently leads to an undesirable concentration of credit risk.

Operational risk has to deal with crimes of “upholstery”, which is a second fraud committed to
cover up the first fraud or mistake. The damage caused by the second fraud generally far ex-
ceeds that of the original loss. Operational risk managers are fortunate that the ability to hide
lo: s constantly shrinking and that the transparency increases over time. A mistake has only
a limited fault tolerance, since different views on the issue help to detect such ones.

Figure 26 clearly summarises the classic overshoot and collapse patterns for each type of risk.
Software maintenance capacity for new software and asset allocation for a holding follow the
same patterns.

Risk Overshoot and Collapse Carrying Capacity
Market Stock market bubbles Price justification
Holdings Asset allocation to holdings
Credit Riskier lenders lend more Creditworthiness
Operational Fraud Hiding capacity
New software Time for maintenance
Business Bank run Cash available

Figure 24: Overshoot and Collapse Structures in Banking

2.4. The Impact of Oscillating Economic Cycles on Risk Management

A series of bubbles in a row becomes a wave or a cycle. Economic conditions like upswings
and downswings result from such dynamics. Investors are mostly not aware of the overall pic-
ture and forget the cyclical nature of markets. These swings are repeated over the long-term
horizon of decades. The daily noise of the market is thereby irrelevant to the macro-cycle. By
reviewing macroeconomic statistics, one can detect the trend and the dynamic impact on bub-
bles. The model includes the GDP variable, which helps us to capture market cycles. Other
cycles depend on the GDP cycle, such as the rating, interest level and stock market cycles.

52 Wikipedia Online Encyclopedia, “Bank Run” article, , available at http://en.wikipedia.org/wiki/Bank_run,
accessed 15 January 2005

53 Example from the System Dynamics Group website: http://www.systemdynamics.org/DL-
IntroSysDyn/start.htm, accessed 11 March 2005.

Figures 27 and 28 depict actual German GDP and modelled GDP:

Figure 25: German GDP Cycles of the Last 40 Years™

Current Angee

Stock price level"

Tiree

attack

GDP

8
5.985
3.97

[One-time terror

1.955 Pl Mutt

-0,06 PI
Interest Level
6

45

3.
1.5

oO
oO 10 20 30 40
Time Yean

Figure 26: GDP from the Model®>

The model uses idealised cycles as in Figure 29. They
do not fully show the actual movements and shapes of
the cycles, but the approximation does include the ebb
and flow of the GDP cycle. None of the cycles from
historical data are exactly the same as the previous
one, because the specific causes of recoveries and
downturns were not the same.*°

54 Datasource: HfB Macroeconomic MasterClass, 2004.

Interest
ral eKe

iStock price

Customer

Npeidults.

Dilation
Deflation
Economic Cycle
Ina
‘Stock Market Cycle

Boor Bul ae

Figure 27: Idealised Cycles®”

55 (range: -0.06 to 8% | with GDP(x)= (2.5*sin((x)/2))+1.5), x=z"n| zeR).
°6 http://www.cornerstoneri.com/home/cycles.htm, accessed 1 March 2005.

57 Ibid.

The GDP cycle is imitated by the following function: GDP(x) = 2.5- sin((x)/2))+1.5, with
X=z-H, x=z-H|zER. Two factors that depend on the GDP cycle are the interest level
curve and the stock price level, which is a bias factor for the stock market index. The GDP
movement impacts the interest spread and the stock market. The standard deviation of each of

the two factors is derived to calculate the economic capital of a bank. A risk model modelled
without the economic cycle would definitely fail to capture these dynamic qualities.

2.5. Time Concerns

Risk levels change over time due to external and internal dynamics. Risk parameters need to be
updated more frequently in times of fast change, since the value function of a risk premium
changes with time. Risk management hunts sometimes phantoms. The perception of senior
bank management changes as knowledge increases. Bank managers might grow impatient and
lose their long-term perspective. Due to advances in technology, the cycle times are shortened.
Straight-through processing (STP) allows for automated loans to customers nearly without hu-
man interaction. On the one hand, STP minimises risk and increases efficiency because it pre-
vents human error. However on the other hand risk might increase through impacts of mistakes
that are more severe, such as wrongly implemented creditworthiness checks or increased cycle
time (i.e. more loans are granted in each period, and a greater number of these will default).

Bank managers look at returns and results more frequently within a shorter time horizon, which
leads to higher risk premiums. This is a myopic loss aversion experience, a narrow view of
results without seeing the GDP cycle or the business environment.* It might also have the ef-
fect that risk managers get used to high volatilities, so that these are seen as normal. The per-
ception of risks varies across countries and among different age groups, so that common stan-
dards have to be set for the measurement of enterprise risk.

Risk Delays
Credit To recognise the debtor’s default.
Market Settlement risk: a position could not be settled during a transaction.
Business Customers are lost because they had to wait too long due to poor service.
Operational To detect fraud and normal mistakes which are invisible.
Technology adoption of the bank itself, of the customers and of the staff.
A non-functioning IT system causes business loss.

Figure 28: Table of Delays in Banking Sorted by Risk Type

Table 30 summarizes the delays appearing in risk management. Delays are fierce and increase
overall bank ri: ince processes are not executed. Delays minimise the quality of bank opera-
tions, because of the formula Risk = /-Quality. Basel II addresses this point in the operational
risk section concerning processes. Risk managers therefore consider delays with regard to IT
systems and try to prevent a situation whereby a trade cannot be executed because an IT server
is down. By no means less severe is a delay in recognising a customer’s default or in detecting
capital misallocation. The latter is particularly dangerous, for the reason that it has “no immedi-
ate impact on company financial statements. However, misallocation will distort ROEs” and
possibly be a catalyst to incorrect decisions being made on strategy.”*°

8 Donald Mango: “Report on CAS Research Working Parties”, Risk Preferences Working Party slide,
available at http://www.casact.org/coneduc/erm/2004/handouts/mango.ppt, accessed 15 January 2005.
°9 ROE = Return on Equity.

6° Vinaya Sharma and Randy Tillis: “Implementing & Transitioning ERM", presentation, slide 25, available
at http://www.casact.org/coneduc/erm/2004/handouts/sharmatillis.ppt, accessed 15 January 2005.

Real-time risk systems permit quicker risk limit checks in the trading room. However, these
systems require data integrity and accuracy. It is necessary to integrate front, middle and back
office in order to perform the required tasks and to reduce the risk of errors due to manual data
input, data re-keying, and data transformation. The question is: how fast can risk information
travel through the whole business to support decision-makers best? A recent article on real-time
analytics questions whether “the immediacy may simply make it easier to make more mistakes,
faster.”*' It was encouraged to implement analytical strategies, which are longer lasting than the
next up sell.

“Once ERM programs are in place, it is important

to make sure the modelling has consistency and validity.”
Mark Puccia (Managing Director - Standard & Poor’s Corp.)

3. The Risk Simulation Model

This chapter shows how the model was developed, what assumptions were made and what in-
sights were gained about risk dynamics. The fundamental steps to build a model are:

= to define a clear-cut purpose for the modelling effort

= to predetermine which factors shall vary and which shall be projected

= to decide, which are the most relevant factors to model.

The modelling process itself as described in figure 31 shows the importance of the problem
definition, since it will determine the meaning for the whole model. Also included in figure 31
are already some results from the findings, which will be discussed in chapter 4.5. in more de-
tail.

Enterprise risk Analyzing enter- Certain risk fac- By observing the
models are only prise risk sys- tors have a strong market volume
toa certain de- temically will influence on the during trades, a
gree dynamic. leads to deeper economic capital. bank can mini-
insight. mize risks.

Identify Develop Test Test Policy

problem Hypothesis Hypothesis Alternatives

Figure 29: The Modelling Process

3.1. Model Development

A dynamic financial model depends on the following four components:
= people available for system design and programming
= data from which to derive assumptions and with which to initialize the model
= money available to purchase an existing software package
= computer architecture.

61 Jason Compton: “Real-time Analytics: Excellent Insight, or Speedier Mistakes?”, available at
http://www.destinationcrm.com/articles/default.asp?ArticleID=4785, accessed 22 January 2005

® Stephen D'Arcy, Richard Gorvett, Joseph Herbers, Thomas Hettinger: “Steps in Building a DFA Model”,
Article available at http://casact.org/coneduc/specsem/98dfa/dfamodel.htm accessed 15 January 2005.

63 System Dynamics: “The Modelling Process”, available at http://www.systemdynamics.org/DL-
IntroSysDyn/start.htm accessed 04 March 2005

64 Dynamic Financial Analysis Committee: “Dynamic Financial Models”, presentation available at
http://casact.org/research/drm/dfahbch2.pdf, page 9, accessed 15 January 2005.

The model was developed by the author from January to March 2005. Thereby, the modelling
process was done iteratively following the Rational Unified Process of software development
(RUP) as shown in figure 32.

Workflows

Business Modeling
Requirements
Analysis & Design

Implementation
Test

Deployment

Configuration
‘& Change Mgmt
Project Management.
Environment

Figure 30: Iterative Development Process according to Rational Unified Process

Phases

Iterations

Requirements

nge’
Management

Test

Anelysis & Design

Implementation

Enviranment Deployment
Evaluation Q

The main data source for the model was the website of Deutsche Bank AG. The data was en-
hanced by trading and finance data from Finance Yahoo, rating agencies and Bloomberg. Some
distributions and benchmark figures were derived from the ERM literature.

Software
components:

Database:

Hardware: Dell
Latitude | 600

Datasources:

Maple 9.5

Modelling process

_ Excel 2000

a.)

Vensim

Rating Website ERM Bloomberg Yahoo HfB Mas-
Agencies Deutsche Literatu- Terminal Finance ter Cour-
Bank AG rey 2 pl 888
Ly

Figure 31: The IT Architecture

To model the GDP and the mathematical formulas, the software Maple was used. This allowed
a flexible calculus, since even sophisticated functions can be plotted easily by Maple. The vari-
ous distributions, their parameters and quantiles were calculated with @Risk developed by the
company Palisade. @Risk is an enhancement to excel. After having calculated the assumptions
and key data, the formulas were entered into Vensim, a simulation software package, which

uses system dynamics notation.

6} Management Systems Consulting: “RUP Fundamentals Presentation”, available at http://www.msc-
inc.net/Documents/rup_fundamentals_presentation.htm, accessed 15 March 2005.

3.2. Basic Assumptions

The following variables were used in the model:

Internal/ endogenous

External / exogenous

Number of assets

Various standard deviations of distributions
Products (number of products, credit, asset)
Staff (number of employees)

Mistakes made by employees

Customers (number, type of, ...)

Pricing (interest rate, year-2000-hype,...)
GDP: We cannot model a bank’s system,
which will not influence the GDP.”
NASDAQ imitated stock price development

Terror Attack Function
Stock Market Crash Function

VaR Values for the single Risk types

The model runs for 40 years in order to integrate the long-term GDP cycles of approximately
10 years. The time steps are measured in years, since the available data of the economic capital
is taken from the annual report. One must keep in mind that economic capital is a notional fig-
ure and that no real money is involved. Therefore, it might be delusive since the economic capi-
tal is measured in monetary terms. The most important risk types are to be modelled, so that the
economic capital as overall risk measure can be calculated.

All the data for the universal bank were taken from the Deutsche Bank. The employee struc-
ture, profit figures and economical capital segmentation imitate the Deutsche Bank figures in a
range of +20%. Deutsche Bank publishes a risk report in each annual report, which was the
main source of information. Where no data were available (e.g. more detailed credit loss provi-
sion calculations or complete probability default rates and distributions), assumptions were
made to imitate the concern at the best.

Figure 35 shows the simplified balance sheet used in the model:

Assets Balance Sheet Liabilities
[in m.€]

Cash 5,000 | Saving Accounts 600,000.
100 Loans 4 500 m€ 500,000 | Current Accounts 40,000
Shares (speculative) 100,000 | Other Liabilities 100,000
Holdings 15,000 | Equity 60,000
Real Estate (5 objects) 100,000. eee
Other Assets 80,000

TOTAL 800,000 | TOTAL 800,000

With a focus on the overall dynamics, flow of single credit payment transactions are aggre-
gated, so that there is no “credit flow”, but a growth rate of credits. The portfolio of the 100
loans with an amount of each 500 m€ were modelled with varying default probabilities (in av-
erage 10%) and a recovery rate of around 30%.

The market risk is determined by three kind of assets:

“NASDAQ-like” shares (standard deviation simulates the historical NASDAQ data)
Holdings, which imitate the historical dax performance

Real estate assets

Historical simulations were made for up to 15 years (1990 to 2005).

66 Compare also John D. Sterman: “A Skeptic's Guide to Computer Models” , 1988/1991, available at
http://www.millenniuminstitute.net/publications/Skeptics.pdf, accessed 04 March 2005, p. 23 et seqq.

Operational risk data were estimated from the portion of economical capital for the operational
risk in the annual report of Deutsche Bank from 2000 to 2005. Benchmark-data from other
international banks were used to enhance the estimation.

3.3. Behavioural Relationships and Initial Conditions

The GDP cycle runs differently than the interest and stock market cycles as explained before.
The stock market influences the shares and holdings, while the interest curve will have an effect
on business revenue and default probabilities.

The indices performance is derived from the long term Dow Jones Industrial average, since it
was the index with the longest data history available. Figure 34 illustrates clearly that between
1995 and 2000 the index accelerated in points and high volume (from 5,000 above 11,000).

DJ INDU AVERAGE (DOW JONES & CO
as of 9-Mar-2005

15000 [7
5000
- 1955_ 1940 is45_1950 i955 15601965 19701975, i980 1s85_1990 1595 2000
5 2op
= tof
°
Copyright 2008 Yahoo! Inc. http://Finance .yahoo .com/

Figure 32: Dow Jones Industrial Average of the last 70 Years

In the model there are certain “IF THEN ELSE” structures implemented regarding two events.
One is a terror attack, which will decrease the GDP over a period of 5 years; the other is a stock
market crash with an exponential recovery over 8 years.

3.4. Tests for Consistency with Purpose and the Boundary

The purpose of the model is to show the dynamics and the driving factors by means of the de-
velopment of the economic capital. The development can be seen for a time period of 40 years
and the effects of a stock market crash and a terror attack can be simulated. The model is con-
sistent for a period of 40 years, although the economical capital skyrockets, which are the effect
of the asset value development. Furthermore, the recovery rate is limited. It cannot exceed
100% and cannot become smaller than 0%. The Economic capital and the VaR figures are not
allowed to become smaller than zero.
3.5. Insights gained from the Model

The recovery rate has an exponential ef-
fect on the economic capital for the credit
risk section. In case banks can recover
with a slightly higher rate, they will be
able to protect assets. The effect from the
default rate is not as strong as the recovery
rate impact. Banks should create an “outer
loop” to credit risk in order to support
riskier lenders by educating companies
beforehand how they can avoid financial
difficulties.

The strongest impact on the economic
capital was the market risk section as it
can be seen on the right side in figure 56.
This big influence is due to the size of the
market position that the imaginary bank
holds. The asset values developed so
strongly that, as a result, the economic
capital increased enormously. To sell
holdings would decrease the economic
capital, but has no major effect compared
with the riskier “NASDAQ position”.

Figure 33: The simulated Results

Operational risk is mainly driven by proc-
ess failures as it can be seen in figure 36.
There is a very high correlation (above
90%) between the development of em-
ployees and the operational economical
capital. A terror attack led even to higher
recoveries in the credit risk, since the GDP
change was small after
the terror attack. Due to

No. of Events

terror a. and s.m.crash
stock market crash
terror attack

Current
base case

Economic Capital

EC Business Risk Inflow
40,000
30,000
20,000
10,000

0

EC Credit Risk Inflow
40,000
30,000
20,000

10,000 == a

0
EC Market Risk Inflow
6M
45M
3M
1.5M
0
EC Operational Risk Inflow
20,000
15,000
10,000
5,000

+ 250
the uncertainty of ter-
rorism, people and cor- 200+— a
porate clients might | ee
become more cautious 150 —— |
while the GDP is at 7
lower rates. 400. Category
504 Process Mgmt.
Fraud
‘ Clients & Products
Employment
Figure 34: Driver Analysis 15275 jai Ansas
of the Operational Risk” 8 750-7 5
750 7.599 7800" 5
500 75,000
Loss ['000 €] 75,000

87 Marcelo Cruz: “Operational Risk Modelling and Analysis”, page 324 Soure: JP Morgan Chase

3.6. Future Extensions of the Model

The model can be extended by distinguishing between corporate and private customers in dif-
ferent countries, since especially concerning credit risk the default probabilities would deviate
from each other very much. With enhanced (internal) data, the exploitation of credit lines could
be modelled. However, the feedback loop “Riskier lenders lend more” is dependent on the data
of customer behaviour. The market risk section can be extended by considering that real estate
market prices depend very much on regional price indices and default probabilities. The time
series input for estimating the real estate standard deviation could be longer (60 years) to cap-
ture whole real estate cycle moves. Not considered were also: yield curve changes in market
tisk. The asset class is to be added in the next version of the model.

The dependence on single risk factor after having considered correlations would be valuable
information to strengthen the risk structure as well. In order to implement this in a future ver-
sion no lookup functions were used. Instead, “IF THEN ELSE” constructions captured the
nonlinearity in the model. An additional questions is what happens in case of a rating downturn:
AA- to B or even to such a bad rating, when no issuing of bank products is possible anymore.
Models already existing in the system dynamics approach could also be used to enhance the
existing model. There have been models for risk assessment of transmission dynamics to calcu-
late how fast epidemics spread among a population." This can be applied to market participants
in an economy, which interact by value chains and “infect” each other with financial difficul-
ties. Figure 37 shows an example. The “New Infect” rate could be calculated as:
Market Participants
Market Participants + Infected + Recovered

Infection Rate = Market Participants*Contact Rate*Transmission Probability*

Companies
alae Deathrates |_Bought
Deathrate Exit Rate for Insolvency | / Pade
i Rate fecovere
‘Market Participants Companies
infected with eb
Fresh Market ge} Financial Recovered
Company | Participants |" Newinfect_| | Distress _| Recoveries Closure Rate of
Formation 4 Sees Business
—— Infection
Contact Rate Duration

Transmission
Probability per
Contact

Figure 35: Example of Transmission of Financial Risk in an Economy

This study would give insight, how and why the “financial epidemic” would rise and fall. How-
ever, this study would need to take a broader look on the economy. It surely would provide
insight about risk concentration as well.

Even a “Risk Management Flight Simulator” could be developed. One cycle or gaming step
would be the semi-annual shareholder conference. Initial conditions are to be shown in terms of
employees, balance sheet, GDP cycle. Then a decision board can be added, which allows to
control the following variables:

= Employee fluctuation (layoff has an impact on service quality and cost base)

= Rating target (from AAA to AA-, a better rating will decrease the cost of capital)

= New product decision (CDO effect is decreasing economic capital)

= Marketing Budget (efforts to work against competition)

68 James Koopman: “Transmission dynamics” available at
http://www.sph.umich.edu/~jkoopman/Web606/RiskAssess/, accessed 16 March 2005

Graphs could illustrate the developments of: P&L, Business Units, Economic Capital, Rating
Tendency, Shareholder value, GDP and Credit Defaults. The gaming procedure of the simu-
lated 10 years could include the following events:

Year _| Event Scenarios
1-2 | Normal mode To learn the market dynamics
3-6 |Competition mode GDP decreases, customer fluctuation up (exponentially),

At a random time point: one merger, a takeover of a suffer-
ing bank and one joint venture of a successful bank

6-8 | Every half-year decision | (1) Creating a CRO position

about one policy to im- _ | (it works out, if the company has the necessary size)
prove the overall risk (2) Implementing a customer education division
structure and EC (will reduce customer defaults, but increases costs)
(3) Selling of holdings

(smaller returns at the end with rising GDP)

(4) Investing more in more riskier assets

(could overstretch the EC in year 8 with low GDP)

9-10 | Normal mode GDP rises again to first level

Such a “risk flight simulation” would certainly increase the learning experience about risk dy-
namics. Future simulations might not only cover single values or risk structures, but whole
companies and processes. Business war games are one such computerised business simulation.
These simulations challenge managers to make decisions that will directly affect their virtual
companies. Over several years, tactical and operational decisions can be tested in a risk-free
simulation environment. This simulation helps managers to understand the dynamics and inter-
actions of their business. They learn faster how to assess the competition accurately, how to
position their company by defining a long-term strategy, and how to make operational cost
decisions.” Forio Business Simulations, for example, offers a price strategy simulator on their
website, showing feedback loops of a price war as depicted in Figure 38.

Revenue Revenue
A

Market Competitor's | Competitor's
Share Price Profits
Sf

Our Profits —> Our Price

SH

Competitor's
Costs

Figure 36: Feedback Relationships in Pricing used by Forio Business Simulations”

Our Costs

69 http://www.prisim.com/, accessed 10 March 2005.
7 Michael Bean: “The Price Strategy Simulator — Anatomy of a Price War", available at
http://www.forio.com/pepricesim.htm, accessed 10 March 2005.

“Do today, what others consider doing tomorrow, only change is constant!”
Heraklit, 480 B.C.

4. Conclusion

ERM focuses on the bank as a system with the bank’s economic capital as notional benchmark
for the overall risk. To provide insight into the correct risk exposure and the robustness of the
bank’s risk structure, the function used to calculate economic capital must be dynamic, as well
as capable of integrating the different risk structures. Credit, market, operational and business
risk differ greatly from each other in terms of uncertainties, default distributions and inherent
potential loss. Only holistic risk management can successfully unite these different types of
risk. Risk management itself is not a new science, but its methods improve constantly. The
focus on ERM requires risk types to be viewed as a portfolio, and allows the bank to implement
an overall risk strategy. However, to gain insight into the risk dynamics and risk interactions, a
dynamic risk model is essential. Complex risk/return decisions can be made only if rapid and
discontinuous changes are considered.

Integrated processes and the vast amount of incomplete information at hand pose a major chal-
lenge to the information architecture of a risk management system. Technology can enhance
risk management capabilities enormously by gathering data and supporting the analytical proc-
ess. Computational power allows more sophisticated scenarios and more accurate modelled
distribution functions, so that a set of different risk measurement techniques can be used to
manage risk. The diversification of the calculus counteracts the systemic risk that all traders
could sell at the same price simply because one benchmark limit has been breached. The en-
hancement of scenarios can even be extended to real world business scenarios such as flight
simulators. This process will become ever more important, as the speed and complexity of the
business decision-making process continues to put pressure on the tolerance limit of banks.
This faster learning cycle helps bank managers to take decisions with enhanced intuitiveness.
Even with the most sophisticated enterprise risk system, a manager will not be able to forecast
chaotic dynamics like bubble behaviours or external shocks.

The value added of the system dynamics method is to increase the robustness of the structure in
terms of non-linear and dynamic change. The right tools will enhance insights gained diagnos-
tically by focusing on the whole bank as a unified system. By doing so, bank managers more
effectively control risk exposures and asset allocations and manage the bank’s organisation as
well as the triangle of risk, return and liquidity within space and time. In a world of rapid
change, dynamic enterprise risk management enables bank managers to see and make proper
use of future opportunities.

Appendix: Model Screenshots

The Model Overview

Credit Risk Module

eae"

Change

DP change” [Prob [aang

} 4
| \Se Loss

Ne
ww Ee
\ Eereiaion -
bE oe

°

Rate of 1 covering
‘Smocthtime —____-—~
\\ hte noise P4

| a

eee
- ~ | _ capi"

The credit risk module cal-
culates the Economic Capi-
tal for a portfolio of 100
loans which an amount of
500 m€ each estimating an
event correlation of 5%.

This module is influenced by
the GDP development. A
thriving business environ-
ment leads to a smaller
number of defaults.

ne,
Market Risk Module

NASDN
eee,

(Growth Rate Decining ~~ re
Nasdaq Nasdaq ‘Standnormimn for the

; 99.98% Quantié 0.0 0

| Stock Crash 0 y -

eg x .
Standard a: peti 4
jetOnos — ee

~*] Economie Capital
for Market Risk |

Real

a a \
Price Growth "RR Pies Déging ‘Standard Pal | | j
Sef
VaR for Real
Seliate of Rea
tte Assets =

Weed
_-¢Standnorminy for the
277 99.98% Quantile 00

The market risk module cal-
culates the Economic Capi-
tal for three assets:

(1) A risky share: simulated
with historical NASDAQ data

y
(2) 5 Real estate assets * 2 Va fo,
| standard — shor )

._ St 2 ;

Sy

»lail (3) Holdings, for which the | “
performance is simulated
with the historical DAX data

‘Standnomminy for the
99.98% Quantile 0

The operational risk module
[Poreniat | simulates fraud, mistakes
a and an external terror at-

oO tack, which is connected to
the GDP development.

(One tine terror

ake a
Employee
Operational Risk ie

‘Standnorminy for the
99.98% Quantile

4
~~ faystem ature
= 0

a
Business Risk

Potertial
Customers
New
~

The business risk module
depends on the number of
customers, which do busi-
ness with the bank and
generate revenue.

) s
\

Y

\V

aeons a ;

‘Standnorminy forthe
99.98% Quantie 1

ya \
ss Starinorirw forthe

99.98% Quartile

References

Monographs and Articles

* Author unknown: “What’s in a Name?”, Journal of Behavioral Finance, Vol. 4, No 1, 2003.

“The Art of Measuring Risk”, Reactions, Vol. 23, Issue 11, 11/2003.

“Fear Factor”, CFO, Vol. 19, Issue 7, 06/2003.

Against the Gods — The Remarkable Story of Risk, Wiley, 1998.

* Caginalp, Gundu; Porter, David and Smith, Vernon L.: “Overreactions, Momentum, Liquidity, &
Price” Journal of Psychology and Financial Markets, Vol. 1/ No 1, 2000.

* Connolly, Jim: “The Newest Buzz at Companies? Risk Management on Steroids”, in: National Under-
writer, 18 October 2004.

* Dowd, Kevin: Beyond Value at Risk, Wiley, 1998.

* Deutsche Bank, Risk Report, in Deutsche Bank Annual Report 2000.

* Deutsche Bank, Risk Report, in Deutsche Bank Annual Report 2001.

* Getmansky, Mila and Papastaikoudi, Jannette: “Extrapolating Expectations: An Explanation for Excess
Volatility and Overreaction”, Scientific Paper, System Dynamics Society.

* HfB Macroeconomic MasterClass, 2004.

* Hockmann, Heinz and ThieBen, Friedrich: Investment Banking, Schafer Poeschel, 2002.

* Kalhoff, Agatha and Haas, Marcus: “Operational Risk - Management Based on the Current Loss Data
Situation”, in: Cruz, Marcelo: Operational Risk Modelling and Analysis — Theory and Practice, Risk
Books, 2004.

* Krell, Eric: “Identify Your Company’s Riskiest Link”, Internet World, Vol. 9 No 5, May 2003.

* Krause, Andreas: “Exploring the Limitations of Value at Risk: How Good Is It in Practice?”, Journal of
Risk Finance, Winter 2003.

* Lam, James: Enterprise Risk Management — From Incentives to Controls, Wiley, 2003.

* Levine, Robert: “Risk Management Systems: Understanding the Need”, Information Systems Man-
agement, Spring 2004.

* Lister, Malcom, SAS: “Intelligent Enterprise Risk Management — Beyond Basel IT”, Internal SAS Pres-
entation on the SAS Conference, 2004.

* Markowitz, Harry M.: “Portfolio Selection”, Journal of Finance, No. 7, 1952.

* Marrison, Chris: The Fundamentals of Risk Measurement, McGraw Hill, 2002.

* McKibben, Douglas: “Banks Evolve in their Approach to ERM”, GartnerG2 Report, June 2004.

* Mussweiler, Thomas and Schneller, Karl: “What Goes Up Must Come Down — How Charts Influence
Decisions to Buy and Sell Stocks”, Journal of Behavioral Finance, Vol. 4, No 3, 2003.

* Rahl, Leslie: “10 Commandments of Risk Management”, Treasury & Risk Management, 07/08 1994.
* Rosenow, Bernd; WeiBbach, Rafael and Altrock, Frank: “Modelling Correlations in Portfolio Credit
Risk” in arXiv:cond-mat/0401329 v1 19.01.2004.

* Renshaw, Edwar: “The Crash of October 19 in Retrospect”, Market Chronicle, 22, 1988.

* Schmidt, Wolfgang: Risk Management, Script Business School of Banking and Finance, 2004.

* Schréck, Gerhard and Steiner, Manfred: “Risk Management and Value Creation in Banks”, in: Risk
Management, Ist edition, Springer, 2004.

* Shapira, Zur: Risk Taking - A Managerial Perspective, Russel Sage, 1995.

* Standard & Poor’s: Rating Performance 2001, Special Report, February 2002.

* Sterman, John D.: “Business Dynamics”, McGraw-Hill, 2000.

* Strohhecker, Jiirgen: “The Object-oriented Model of a System’s Dynamic”, in: System- und objektori-
entierte Simulation betriebswirtschaftlicher Entscheidungen, Scientific Papers of the Industrieseminar
University Mannheim, Issue 53, Berlin: Duncker & Humblot, 1998.

* Taleb, Nassim: Fooled by Randomness, 2nd edition, Texere, 2004.

* Wiechers, Riidiger: “Vertrieb und Risiko — Vertrieb als Risiko“, Immobilien & Finanzierung — Der
langfristige Kredit, No 11, 01 June 2004.

Internet Websites

Jootar, Jay: “A Risk Dynamics Model of Complex System Development”,
http://theses.mit.edu/Dienst/UI/2.0/ShowPage/0018.mit.theses%2f2002-
227?npages=204&format=inline&page=1, accessed 10 March 2005.

Toma’s Report on Financial Analysis, Planning & Reporting, www.ioma.com, 12/2004.

Saunders, John: “A Dynamic Risk Model for InformationTechnology Security”,

http://www. johnsaunders.com/papers/riskcip/RiskConference.htm, accessed 15 January 2005.
http://www.cogsci.princeton.edu/cgi-bin/webwn2.0?stage=1 &word=risk, accessed 28 January 2005.
http://econ.worldbank.org/view.php?id=39805, accessed 4 March 2005.
http://65.118.47.249/riskmanagepart4.html, accessed 16 March 2005.
http://www.bis.org/publ/bcbsca.htm, accessed 25 February 2005.

http://systemdynamics.org/, accessed 25 February 2005.
http://www.systemdynamics.org/DL-IntroSysDyn/start.htm, accessed 16 March 2005.
http://riskmanager.blogspot.com, accessed 27 February 2005.

Centre for the Study of Financial Innovation and PWC: Press Release, 21 February 2005,
http://www.a2mediagroup.com/?c=47 &a=214 1 &sid=127bc9d6e3828deafeb59607 13f086c6, and
http://www.csfi.org.uk/; accessed 27 February 2005.

Skora, Ricard K.: “Modern Portfolio Credit Risk Modeling”, http://www.skora.com/modern.pdf, ac-
cessed 6 March 2005.

http://www.bankenverband.de/pic/artikelpic/122004/78 |-Kredit-BGw-10-04.pdf, accessed 28 February
2005.

http://finance. yahoo.com/q/hp?s=%5EIXIC&a=00&b=1 &c=1984&d=01 &e=25&f=2005&g=d, accessed
25 February 2005.

BIS, http://www.bis.org/publ/bcbsca.htm.

http://www.fmsinc.org/cms/?pid=1001427, accessed 28 February 2005.
http://www.chicagofed.org/banking_information/legal_reputational_risk.cfm, accessed 3 February 2005.
http://www.defaultrisk.com/pp_super_41.htm, accessed 27 February 2005.
http://www.csfb.com/institutional/research/CreditPortfolioModeling.pdf, accessed 1 March 2005.

Rahl, Leslie: Lessons learned, Risk Insights, 01/2002, www.cmra.com

Brander, Jim and Manoff, Sam: “ERM and DFA Using Active Knowledge Structures”,
http://www.casact.org/pubs/forum/03sforum/03sf001.pdf; accessed 15 January 2005.

Carey, Mark: “Changed World, New Risks”, presentation in 2003, slide 18,
http://www.delcreo.com/delcreo/free/docs/Changed%20W orld%2C%20New%20Risks.ppt, accessed 7
January 2005.

http://www.duke.edu/~mjd/chaos/chaosh.html, accessed 27 February 2005.

Wikipedia Online Encyclopedia, http://en.wikipedia.org/wiki/Chaos_theory, accessed 27 February 2005.
https://royercenter.cwe.psu.edu/Homes/jds100/john/kbase2/systems/page_61.html, accessed 27 February
2005.

http://whatis.techtarget.com/definition/0,,sid9_gci759332,00.html, accessed 26 February 2005.
http://www. greatdreams.com/buterfly.htm, accessed 26 February 2005.

http://www.sasked. gov.sk.ca/docs/chemistry/ksc_b.html, accessed 3 February 2005.

Martin, Leslie: “An Introduction to Feedback”, 1997,
http://sysdyn.clexchange.org/sdep/Roadmaps/RM2/D-469 L pdf, accessed 11 March 2005.
http://www.systemdynamics.org/DL-IntroSysDyn/start.htm, accessed 11 March 2005.

Frye, John: “Loss Given Default and Credit Portfolio Risk”, presentation at the Symposium on Enterprise
Wide Risk Management in Chicago, 26 April 2004,
http://www.casact.org/coneduc/erm/2004/handouts/fryebw.ppt, accessed 27 January 2005.

Intel Research Website, http://www. intel.com/research/silicon/mooreslaw.htm, accessed11 March 2005.
Zhu, Helen: “Mental Simulation of Simple Negative Feedback”,
http://sysdyn.clexchange.org/sdep/Roadmaps/RM3/D-4536-2.pdf, accessed 12 March 2005.

Wikipedia Online Encyclopedia, http://en.wikipedia.org/wiki/Soap_bubble, accessed 24 February 2005.
Miillenmeister, Jorg, http://www. wallstreet-

online.de/ws/community/board/threadpages. php? &tid=00771850&reverse=1 &page=8, accessed 23 Feb-
ruary 2005.

http://www.caslon.com.au/boomprofile 1.htm#railway, accessed 4 March 2005.
http://www.investopedia.com/features/crashes/crashes|.asp, accessed 4 March 2005.
http://www.cornerstoneri.com/newpage17.htm, accessed 1 March 2005.

Wolfram Research Inc., http://mathworld.wolfram.com/Sphere.html, accessed 23 February 2005.
Camerer, Colin: “Market Efficiency of Bubbles”, Journal of Psychology and Financial Markets, Vol. 3,
No 1, 2002.

Breierova, Lucia: “Generic Structures: Overshoot and Collapse”, 1997,
http://sysdyn.clexchange.org/sdep/Roadmaps/RM9/D-4480.pdf, accessed 12 March 2005.

Wikipedia Online Encyclopedia, “Bank Run” article, http://en.wikipedia.org/wiki/Bank_run, accessed 15
January 2005.

http://www.systemdynamics.org/DL-IntroSysDyn/start.htm, accessed 11 March 2005.
http://www.cornerstoneri.com/home/cycles.htm, accessed 1 March 2005.

Mango, Donald: “Report on CAS Research Working Parties”, Risk Preferences Working Party slide,
http://www.casact.org/coneduc/erm/2004/handouts/mango.ppt, accessed 15 January 2005.

Sharma, Vinaya and Tillis, Randy: “Implementing & Transitioning ERM”, presentation, slide 25,
http://www.casact.org/coneduc/erm/2004/handouts/sharmatillis.ppt, accessed 15 January 2005.
D’Arcy, Stephen; Gorvett, Richard; Herbers, Joseph and Hettinger, Thomas: “Steps in Building a DFA
Model”, http://casact.org/coneduc/specsem/98dfa/dfamodel.htm, Dynamic Financial Analysis Commit-
tee: “Dynamic Financial Models”, accessed 15 January 2005.
http://casact.org/research/drm/dfahbch2.pdf

Management Systems Consulting: “RUP Fundamentals Presentation”, http://www.msc-
inc.net/Documents/rup_fundamentals_presentation.htm, accessed 15 March 2005.

Sterman, John D.: “A Skeptic’s Guide to Computer Models”, 1988/1991,
http://www.millenniuminstitute.net/publications/Skeptics.pdf, accessed 04 March 2005.

Koopman, James: “Transmission dynamics”,
http://www.sph.umich.edu/~jkoopman/Web606/RiskAssess/, accessed 16 March 2005.

Glass-Husain, Will: “The Drive to learn” http://www.forio.com/article_motivation.htm, accessed 10
March 2005.

Tillis, Randy, “Enterprise Risk Management”, presentation, slide 10,
http://www.casact.org/coneduc/erm/2004/handouts/tillis2.ppt, accessed 15 January 2005.
http://www.systemdynamics.org/DL-IntroSysDyn/start.htm, accessed 13 March 2005.

http://www. financetech.com/focus/standards/showArticle.jhtml?articleID=54200152, accessed 27 Febru-
ary 2005.

Fitch Ratings,

http://www. fitchratings.com/corporate/reports/report_frame.cfm?rpt_id=226312&sector_flag=1&market
sector=6, accessed 28 February 2005.

http://www. isda.org/statistics/, accessed 28 February 2005.

FRBC: “Risk Committee Market Risk Observations”,
http://www.chicagofed.org/banking_information/market_risk.cfm, accessed 13 March 2005.

Wolff, Ken: “Momentum Trading: Calling a Halt to Missing Stops”, 14 February 2000,
http://www.thestreet.com/funds/onlineinvesting/883534.html, accessed 27 December 2004.

The handbook of world stock, derivative and commodity exchange https://www.exchange-
handbook.co.uk/news_story.cfm?id=50750, accessed 23 January 2005.
http://www.day-trading.de/Beginner/CircuitBreaker.php, accessed 29 December 2004.
http://changingminds.org/explanations/theories/hindsight_bias.htm, accessed 4 March 2005.

FRBC: “Operational Risk”, http://www.chicagofed.org/banking_information/operational_risk.cfm, ac-
cessed 20 February 2005.

Compton, Jason: “Real-time Analytics: Excellent Insight, or Speedier Mistakes?”,
http://www.destinationcrm.com/articles/default.asp?ArticleID=4785, accessed 22 January 2005.
“Bank Governance Principles”, http://www.apec-finsecreg.org/BankGovernance.htm, accessed 13 March
2005.

Breeden, Joseph: “Stress-testing”, September 2004,
http://www.strategicanalytics.com/topics/2004rmaj_stresstesting.php, accessed 13 March 2005.
Johnson, Neil: presentation, http://www.occf.ox.ac.uk/bin/pres_small/sld019.htm, accessed 13 March
2005.

Brander, Jim and Manoff,Sam: “ERM and DFA Using Active Knowledge Structures”, paper,
http://www.casact.org/pubs/forum/03sforum/03sf001.pdf; accessed 15 January 2005.
http://www.prisim.com/, accessed 10 March 2005.

Bean, Michael: “The Price Strategy Simulator — Anatomy of a Price War”,
http://www.forio.com/pcpricesim.htm, accessed 10 March 2005.

German Advisory Council on Global Change http://www.wbgu.de/wbgu_jg1998_kurz_engl.html, ac-
cessed 16 March 2005.

Lam, James: “ERM: Beyond Regulatory and Governance Standards”, July 2004,

http://www. isda.org/c_and_a/ppt/James-Lam-July23-04.ppt, accessed 14 March 2005.

Federal Reserve Bank of Chicago: “Operational Risk”
http://www.chicagofed.org/banking_information/operational_risk.cfm, accessed 20 February 2005.
http://www.systemdynamics.org/DL-IntroSysDyn/start.htm, accessed 12 March 2005.

.@ model bank\@Risk Structures and Drivers\own\2005-03-17 whole risk moc

Fri Mar 18, 2005 5:33AM