Capco JournAL PDF
Capco JournAL PDF
#38
09.2013
PUTTING IT ALL
TOGETHER FOR
OUR CLIENTS ...
Thats what makes FIS the global leader.
www.fisglobal.com
2012 FIS and/or its subsidiaries. All Rights Reserved.
Journal
The Capco Institute Journal of Financial Transformation
Editor
Prof. Damiano Brigo, Director of the Capco Institute and Head of the Mathematical
Finance Research Group, Imperial College, London
Advisory Editors
Cornel Bender, Partner, Capco
Nick Jackson, Partner, Capco
Editorial Board
Franklin Allen, Nippon Life Professor of Finance, The Wharton School,
University of Pennsylvania
Joe Anastasio, Partner, Capco
Philippe dArvisenet, Group Chief Economist, BNP Paribas
Rudi Bogni, former Chief Executive Officer, UBS Private Banking
Bruno Bonati, Strategic Consultant, Bruno Bonati Consulting
David Clark, NED on the board of financial institutions and a former senior
advisor to the FSA
Gry Daeninck, former CEO, Robeco
Stephen C. Daffron, former Global Head of Operations, Morgan Stanley
Douglas W. Diamond, Merton H. Miller Distinguished Service Professor of Finance,
Graduate School of Business, University of Chicago
Elroy Dimson, Professor Emeritus, London Business School
Nicholas Economides, Professor of Economics, Leonard N. Stern School of
Business, New York University
Michael Enthoven, Former Chief Executive Officer, NIBC Bank N.V.
Jos Luis Escriv, Group Chief Economist, Grupo BBVA
George Feiger, Executive Vice President and Head of Wealth Management,
Zions Bancorporation
Gregorio de Felice, Group Chief Economist, Banca Intesa
Hans Geiger, Professor of Banking, Swiss Banking Institute, University of Zurich
Peter Gomber, Full Professor, Chair of e-Finance, Goethe University Frankfurt
Wilfried Hauck, Chief Executive Officer, Allianz Dresdner Asset Management
International GmbH
Pierre Hillion, de Picciotto Chaired Professor of Alternative Investments and
Shell Professor of Finance, INSEAD
Thomas Kloet, Chief Executive Officer, TMX Group Inc.
Mitchel Lenson, former Group Head of IT and Operations, Deutsche Bank Group
Donald A. Marchand, Professor of Strategy and Information Management,
IMD and Chairman and President of enterpriseIQ
Colin Mayer, Peter Moores Dean, Sad Business School, Oxford University
Steve Perry, Executive Vice President, Visa Europe
Derek Sach, Head of Global Restructuring, The Royal Bank of Scotland
ManMohan S. Sodhi, Professor in Operations & Supply Chain Management,
Cass Business School, City University London
John Taysom, Founder & Joint CEO, The Reuters Greenhouse Fund
Graham Vickery, Head of Information Economy Unit, OECD
Zicklin-Capco
Institute Paper Series
in Applied Finance
Ideas at Work
Cutting Edge
67
83
High-level Debate
17
35
41
47
55
Dear Reader,
Time flies. It only seems like yesterday
optimism is returning.
Rob Heyvaert
Expanding Finance:
Conditions at the Boundary
At the present time, when we are still try-
valuing brands.
to welcome submissions.
College, London
Ideas
at
Work
Ideas
at
Work
A Journey to
Operational Excellence
Julien Blanchet, Managing Principal, Capco
Stphane Arvor, Principal Consultant, Capco
Bethsabe Fitoussi, Principal Consultant, Capco
Galle Laboureur, Senior Consultant, Capco
Emmanuelle Mayet-Delors, Consultant, Capco
Introduction
Transform management culture. Employees at large banks are impacted by the rapid
Conviction is everything
Methodologies
where.
Basic principles
Doing
of continuous improvement.
company culture.
Thinking
Performance management;
Cultural change.
people.
and experience. It takes research and out-ofthe-box thinking to define new practices and
Increase in turnover;
phased approaches.
agement.
higher performance.
Ideas
at
Work
Lean
Six Sigma
Focuses on process
to eliminate non-value
added activities and
improve velocity
Focuses on variance
to remove defects and
improve quality and
reliability
Provides tools to
analyze flow and delay
time
Provides statistical
techniques and other
tools to identify root
causes of defects
Lean Management
on a daily basis.
client:
improvement mechanisms.
process efficiency;
The pilot
We agreed to deploy a six-month Prove the
sustainability.
satisfaction.
10
provement mechanisms.
to core process.
Quantitative gains:
Significant RWA reduction through better coordination between risk and finance on the
Basel II process
ager).
Qualitative gains
The certification of seven Green Belts and the mobilization of numerous stakeholders
cesses took place at the right level. With proThis success story led to the creation of an operational efficiency program sponsored by the
Building up capacity
Champion
Role
Coordinates program
Links sponsors and senior
management
Reports on program progress
Master
Black Belt
Black Belts
Green Belts
Yellow Belts
11
Ideas
at
Work
an operational risk analysis (FMEA4) was conducted to identify high-risk process steps and
Strong encouragement of the team leaders and operational staff to participate in the
realized.
objectives.
three phases:
sion of its geographical perimeter, the operational team was facing an increasing number
12
change.
phases is the alignment of the whole managing team concerning what needs to be done
and why.
one. It is a pragmatic approach where the difficulty does not lie in the concepts but in the
Conclusion
Your journey to operational excellence will not
start unless you decide to take it. Banks believing that spontaneous momentum is all that
is required will never make it. But once you
take the decision you can kick off safe in the
knowledge that there are methodologies with
a proven track record in the financial industry.
In order to increase the chances of achieving
your objectives, your operational efficiency
program team will combine various methodologies and project approaches that make it
possible to adapt effectively to your specific
challenges and constraints. All methodologies
will address process efficiency and simplification, optimization of team management, and
continuous improvement. However, a methodology alone is not sufficient. The operational
efficiency team will need to combine it with
in-depth knowledge and experience of the
banks business processes.
Results are based on well-known tools but
their effective implementation depends on the
proper prioritization of opportunities and the
skills of the operational efficiency team. When
13
High-level Debate
16
High-level Debate
Enhancing Islamic
Finance through
Risk Benchmarking
Rodrigo Zepeda Independent Consultant and Associate, The Chartered Institute for
Securities & Investment
Abstract
they are more equitable, stable and ethical, and offer less
17
Introduction
The most recent estimates point to global Islamic finance assets reaching
approximately US$1.8 trillion by 2013, with the top 20 Islamic banks reg-
the world, with over 1.6 billion Muslim followers of Islam. For Muslims, Is-
istering a growth of 16% over the last three years, while holding over 50%
lam means submission to the will of Allah, who is believed to have mani-
of Islamic banking assets globally [Ernst & Young (2012)]. Figures from
the 7th century CE. These recitations were revealed to the Prophet Mu-
hammad (pbuh) during his lifetime by the angel Gabriel (Jibril), and were
as Malaysia and the Middle East (Egypt, Iran, Kuwait, Qatar, Saudi Arabia)
have taken center stage in terms of asset levels, there are new markets
on the horizon such as Iraq and Indonesia [The Banker (2008)]. Added
to this, the Islamic finance industry has registered more than 20% an-
nual growth since 2000, and with Muslims accounting for approximately
20% of the global population, the potential for future growth worldwide is
system which is derived from both primary and secondary sources (Usul
al-Fiqh). The primary sources are the Quran (the divine scripture), the
ing to lag behind that of conventional Western banking in the same mar-
mad (pbuh)), and the Hadith (the acts or sayings ascribed to the Prophet
Muhammad (pbuh) and written down). The secondary sources are for-
propositions; and a very basic operational risk culture [Ernst & Young
(2012)]. In this new post-financial crisis era, it has been argued that firms
ship of risks, the development of a true risk culture, and improved trans-
the same time it has been observed that many Islamic BFIs based in the
Middle East (i.e., Bahrain, Lebanon, Saudi Arabia, Qatar, the United Arab
Emirates (UAE)) lack risk management systems or software, and are lag-
shariaa govern every aspect of a Muslims life, including the way he con-
ducts his business, the criteria for valid contracts and the prohibitions.
ment practices [Deloitte (2010), (2011)]. This paper will therefore argue
for public and private Islamic BFIs operating regionally (excluding central
banks) would contribute to the enhancement and development of Islamic
or contracts are valid or permissible (halal) unless they are either forbid-
finance products. This paper will therefore provide a high level analysis
den (haram), or fall under the purview of any one of the aforementioned
of the theoretical framework within which the new Islamic finance risk
benchmarking could operate in practice. The paper will first explain what
which involve inter alia pork (or pork products); alcohol; armaments;
products offered. Next the paper will discuss financial risk and financial
(i) riba (essentially interest); (ii) gharar (uncertainty); or (iii) maisir or qimr
tice. Finally, the paper will apply these ideas with a view to developing a
guaranteed where the rate of return is tied to the maturity and principal,
rather than the performance, of an investment [Mohammed and Kianfar
18
from the Quran which proscribe the use of riba.1 The principle is also
based on the fact that Islam does not recognize the intrinsic or time-
involve the buyer identifying a commodity or goods (e.g., a car) which the
Islamic bank will purchase at the behest of the buyer, and then sell on to
land or labor, rather than through the lending of money, as well as deriv-
the buyer at a fixed rate of profit, i.e., cost-plus sale. The cost price of
the commodity is disclosed to the buyer and the profit element reflects
the risk (mukhatarahof) of ownership and business risk which the Islamic
bank must take on prior to the re-sale of the commodity (as no increase in
the mark-up is allowed if the buyer does not take up the goods). However,
ment basis through the use of a Shariah compliant credit sale (Muraba-
should not lack specificity of terms (e.g., sale price, deliverability, quantity,
the Islamic bank would purchase the commodity identified by the buyer,
and would then re-sell the goods to the buyer on a cost-plus basis with
sale is Shariah compliant so long as the period provided for future repay-
mon practices include the purchase and sale of goods that never move
from the warehouse, e.g., construction materials. It has also been noted
that certain Islamic banks have specific space allocated to them in mer-
ing on in the wake of the recent global financial crisis, commercial activities
sible may now instead be regarded as falling foul of the gharar or maisir
it has become very widely used for trade financing purposes, however
Islamic banks have to face additional asset risk, greater fiduciary risks,
greater legal risk and the Sharah compliance risk [Ayub (2008), 233].
a gain from some productive effort. However, it is invariably the case that
the degree of speculation and the subjective intention will be interpreted in
two parties whereby one party, which is typically the Islamic bank, will act
as the investor (rabb almal or rabbul-ml), and the other party will provide
stance, for some Islamic finance scholars it is believed that certain types
1 For example, Surah al-Rum, 30 verse 39 states: That which you give as Riba to increase
the peoples wealth increases not with God; but that which you give in charity, seeking
the goodwill of God, multiplies manifold. Surah al-Nisa, 4 verse 161 states: And for their
taking Riba although it was forbidden for them, and their wrongful appropriation of other
peoples property. We have prepared for those among them who reject faith a grievous
punishment. Surah al-Baqarah, verse 275 states: Those who take Riba shall be raised
like those who have been driven to madness by the touch of the Devil; this is because
they say: Trade is just like interest while God has permitted trade and forbidden interest.
Hence those who have received the admonition from their Lord and desist, may keep
their previous gains, their case being entrusted to God; but those who revert, shall be the
inhabitants of the fire and abide therein forever [Ayub (2008), 45].
2 For example, Surah al-Maidah, 5 verse 90 states: O you who believe! Intoxicants and
gambling, sacrificing to stones, and divination by arrows, are abominable actions of Satan;
so abstain from them, that you may prosper. [Ayub (2008), 62].
19
the contract. Any profits that arise from the joint venture will be shared by
Sukuk are medium (3-10 year) to long term (10+ years) Shariah com-
bears the risk of loss. If more than one investor is investing in the Mudaraba
way as equity financing ventures (e.g., seed financing), however there are
two important differences. Firstly, the venture must be made in a halal ac-
resent a profit and risk sharing agreement between the issuer and the
tivity or underlying asset. This means that for example, a companys core
such as alcohol (e.g., bars) or gambling (e.g., casinos) then this would be
ing that for example, so long as the haram activity accounts for no more
then that will not constitute a core business and will be allowed under
(transfer of actual ownership or true sale), but all sukuk usually involve
Shariah rules. However, this rule has not been applied consistently across
all Islamic finance jurisdictions and SSBs, meaning that in practice the ex-
sukuk are generally more risky to Islamic investors in terms of the oc-
noted that the final distribution of profits under a Mudaraba can only take
assets (as well as share sale proceeds of the assets upon liquidation),
The Musharaka (or Musharakah) contract functions in much the same way
tinguished from the Mudaraba in that all parties may contribute financial
investments in the joint venture, and they may all take part in its manage-
assets. This impacts upon the risk profile of Islamic sukuk instruments,
meaning that while AAA rated Western long-term bonds are seen as low
vance as this would preclude risk taking. Furthermore, any losses made in
ary markets for sukuk issues, together with the increased financial risk
they are regularly used to fund the purchase of property through home pur-
chase plans. Essentially, these involve a property being divided into units
and one party to the contract gradually buying all the units from the other
or cash flows has proven difficult, and consequently recent efforts have
party over a fixed period (e.g., 15 years), at which point full ownership of
seldom used because of the high degree of uncertainty over the partner-
market is not very active. Although quotes are provided by some market
ship returns, and thus it is predominantly used in cases which involve large
makers, the spreads between bid and asking price are particularly wide
institutions major risks. This is because they argue that peoples tenden-
finance for both operational lease (ijara), or finance lease (with purchase
the onus is on the institution to identify the few risks that really matter,
which is typically four to six key risks most commonly encountered (i.e.,
asset subject to lease must continue to hold value throughout the dura-
such as oil); and the lessor must hold the risk of ownership (i.e., liability for
company is the natural owner of the risk involved, since natural owner-
maintenance and insurance) throughout the term of the lease. Also, if the
ship assessment yields a clear corporate risk strategy which can create
leased asset is destroyed or the lessor is prevented from using the leased
asset then the ijara contract is terminated and the parties are released from
their obligations under the lease. Chinoy (1995) states that in order to con-
form to Shariah the ijara is based on profit sharing, so that rentals consti-
al. (2008)]. In fact, Gerken et al. (2010) believe that getting risk ownership
tuting the Islamic banks return are not calculated on the basis of capital
recent years many firms owned risks that they either had not considered
by the lessee, i.e., rental payments are a fixed percentage of annual profits
or for which they were ill suited. It is therefore argued that what is needed
attributable to leased asset output. Chinoy (1995) further adds that Islamic
is that an institution takes on board only those risks for which it has three
ijara leasing has the potential for Shariah securitization (under a Shariah
core lines of defense, namely (i) a resilient and flexible business model;
certificate of debt) because the funding can be totally matched from a li-
(ii) well-developed risk skills and capabilities; and (iii) sufficient financial
quidity perspective to the asset being purchase, and the cash flow is rela-
tively certain. In practice ijara leases are commonly used as part of Islamic
on the three major risk groupings faced by all major banks and credit
of the mortgage, make lease and capital repayments throughout, and title
institutions, namely market risk, credit risk, and operational risk [Ioan-
Within the fields of economics and finance there exist a plethora of both
Market risk
well known and less familiar risk typologies. These include, but are not
limited to, acquisition risk; commodity risk; country risk; credit risk; di-
saster recovery risk; environment risk; equity market risk; inflation risk;
interest rate risk; investment risk; legal risk; liquidity risk; operational risk;
book, and forex and commodities risks throughout the bank [BIS (2006),
risk; and volatility risk. In practice, certain types of risk can be signifi-
157]. Market risk may be comprised of sub-risks which make up the total
market risk faced by a bank or financial institution. It will also reflect the
trading), and the nature of the business undertaken will in turn affect the
of risk such as systemic risk, which may affect the interlinkages and
weighting of any sub-risks faced. For instance, sub-risks may include cur-
rency risk (risk of variation of forex rates); commodity risk (risk of variation
prices); interest rate risk (risk of variation of interest rates); and liquid-
ity risk (risk that the price at which you buy (or sell) something may be
significantly less advantageous than the price you could have achieved
under more ideal conditions [Allen (2013), 3]). Thus, where financial
This can be done using the same methodology as would be required for
In this regard Buehler et al (2008) argue that the first step in developing
it is stated that credit and financial institutions now widely use value-
21
over a specified period (often the next day or next 10 days) that will be
level of 1%) for internal risk management purposes (i.e., maximum loss
have been developed and used which rely on more objective methodolo-
level) [Ioannis et al. (2006), 1-2]. VaR is said to be precisely derived from a
gies to modeling credit risk, including the structural approach, the re-
duced form approach, and the actuarial based approach. The struc-
(P) over time horizon (T) (Figure 1) [Medova and Kyriacou (2001)].
There are three further points to be noted about VaR. The first is that it
is derived from the Capital Asset Pricing Model (CAPM), which is itself
ing competitively and all using risk-free rate, single period time horizon,
asset-based rates of return using equity returns (as asset returns are un-
[Jarrow and Turnball (2000)]. The reduced form model of credit default
ity [Allen et al. (2003)]). The third point is there are different quantitative
methods for calculating VaR and the method used is dependent on the
portfolio type and the data available for modeling, e.g., variance-co-
of latent state variables [Arora et al. (2005)]. Default probabilities are nor-
historical default data [Jarrow and Turnball (1995), Hull and White (2000)].
Arora et al. (2005) believe that while this type of model does not com-
Credit risk
that there is a lack of clear economic rationale for defining the nature of
credit transaction will fail to meet its payment obligations under the terms
models credit risk as default loss using expected default rate risk drivers,
and using a function of shared risk factors such as default rates and vola-
tions liquidity, which represents a real and significant risk to banks, since
more than 70% of a banks balance sheet relates to credit risk manage-
relevant factors such as the portfolio type, obligor profile, credit policies,
ment [van Greuning and Iqbal (2008)]. A widely used source of measuring
Operational risk
Moodys, or Standard & Poors. However, these types of credit ratings are
Operational risk is defined by the BIS as the risk of loss resulting from
inadequate or failed internal processes, people and systems or from external events, and while legal risk3 is included within the definition, stra-
Pr[P(T) -VaR]=
F-1 () = VaR (m=0)
tegic and reputational risk are excluded [BIS (2006), 144]. Operational risk
may include sub-risks such as accounting risk; enterprise risk; funding
liquidity risk; legal risk; operations risk; and technology and systems risk.
The chief difficulty with identifying operational risk is its extensive scope
and lack of homogeneity among different BFIs. For instance, people risks
may be dependent on the size of the firm, and technology risks may
be dependent on the firms technology investment and infrastructure. Indeed Akkizidis and Khandelwal (2008) have noted that the wide range
VaR
22
3 The BIS states: Legal risk includes, but is not limited to, exposure to fines, penalties, or
punitive damages resulting from supervisory actions, as well as private settlements [BIS
(2006), 144].
theless, BFIs have been developing models which rely on a similar set
From one perspective it has been argued that operational risk is usually
defined in the negative (as it includes all risks which are not categorized
as either market or credit risks), but that in any event it comprises virtu-
ally all the risk that cannot be managed through the use of liquid mar-
ing against Islam, and would in principle breach his or her trust with Al-
cation [BIS (2006)]. These methods (the Basic Indicator Approach, the
equate capital charges with either annual gross income (previous three
quantitative operational risk models [BIS (2006)]. The BIS approach how-
apply to Islamic institutions that fall under the purview of any particular
ever, does not include reputational risk which is a significant risk in the Is-
one SSB may not be accepted by other SSBs acting in different jurisdic-
risk modeling approaches which have also been developed (Table 1: Ap-
(IFSB) and The Accounting and Auditing Organization for Islamic Finan-
compliance is sine qua non of Islamic finance. In short Shariah risk can
be simply defined as the risk that arises from a failure to comply with
that Shariah compliance does not automatically entail the ethical and
Shariah rules and principles, and applies just as much to Islamic BFIs as
economic growth, so that input and output legitimacy are often not fully
it as the risk that the terms in the agreed in a contract do not effectively
congruent. Nevertheless, Shariah risk still carries with it not only person-
comply with Islamic jurisprudence and thus are not valid under Islamic
al repercussions for the Muslim customer, but also reputational repercussions for the offending Islamic bank or financial institution, since breach
of Shariah rules may cause significant reputational harm and censure
Process approaches
Factor approaches
Actuarial approaches
Causal
Risk indicators
lar jurisdictions and for Islamic institutions operating under the guidance
CAPM-like models
Predictive models
of any particular SSB. For instance, this may translate in practice for
4 An-Nahl (16 Verse 27): Then, on the Day of Resurrection, He will disgrace them and will
say: Where are My (so called) partners concerning whom you used to disagree and
dispute (with the believers, by defying and disobeying Allah)? Those who have been given
the knowledge (about the Torment of Allah for the disbelievers) will say: Verily! Disgrace
this Day and misery are upon the disbelievers.
23
Classification
Type
Meaning
Nature of
relevant other
Internal
Competitor
Industry
Generic
Global
Process
Functional
(leverage) is no more than 10-33%, and total illiquid assets are less than
10-33% of its total assets).
Content of
benchmarking
Camp (1989): Benchmarking is the search for the best industry practices
which will lead to exceptional performance through the implementation of
these best practices. Delpachitra and Beal (2002) opine that while modern
benchmarking involves comparing the performance of organizations, the
primary objective of benchmarking is to increase the probability of attaining competitive advantage. On the other hand, Anand and Kodali (2008)
define benchmarking as a continuous analysis of strategies, functions,
processes, products or services, performances, etc. compared within or
between best-in-class organizations by obtaining information through ap-
Performance
Strategic
Competitive
Collaborative
propriate data collection method, with the intention of assessing an organisations current standards and thereby carrying out self-improvement by
In light of this Moriarty (2011, 598) argues that [a] benchmarking process
functional, generic, operational, performance, process, product, and strategic benchmarking typologies (Table 2).
24
might, for example, involve the use of financial analysis to compare and
the last decade [Mukherjee et al. (2002)]. Mukherjee et al. (2002) under-
(KPIs) or other metrics such as gross, operating, and net profit margins;
to fixed assets ratios; revenue or cost per employee; or sales and profit-
ability trends. From the outset it should be noted that what is important
Performance was defined as how a bank was able to utilize its resources
was interesting about the study was that the DEA clustering highlighted
poor-performing clusters, enabling the identification of weak performing
banks, which were subsequently probed and found to have underlying
Planning
assets, high costs of funds, and low capital adequacy ratios [Mukherjee
et al. (2002)].
Analysis
This highlights the point that the importance of benchmarking does not
simply lie in the process itself, but rather in the final interpretation of the
6. Communicate findings and gain acceptance
Integration
7. Establish functional goals
management? However, at the same time it has also been argued that
criticisms of benchmarking include that it is naturally an a posteriori or-
Action
also be argued that there are inherent difficulties peculiar to benchmarking, including the accuracy and availability of comparator data or factors,
since the most apposite benchmarking may sometimes require obtain-
ing valuable competitor market data. Yet this contradicts one of the fundamental underlying objectives behind benchmarking, namely attaining
effort, external advisors, and costs); (ii) processes (i.e., risk identifica-
event horizon scanning); (vi) monitoring (i.e., assessing payback); and (vii)
yield, for the embedded market and credit risks [Standard and Poors
etary Risk-to-Price (R2P) scores for corporate debt issues through the
spread [Standard and Poors (2013)]. This R2P risk benchmarking typol-
(Figure 2).
cial risk benchmarking there are certainly recognized procedures, frameworks, and inputs that can be utilized. The difficulty in practice however,
Thus Harry Lipman, the Global Product Manager for Bloomberg OTC
payments on, and set prices for, at least US$750 trillion worth of finan-
since risk metrics are still relatively in their infancy, much work may need
level playing field for all, and furthermore that they should be acces-
25
95%
93%
90%
90%
79%
72%
65%
59%
59%
50%
43%
41%
24%
41%
0%
Series 1
17%
13%
5%
Governance
26
Quality of the
benchmark
Quality of the
methodology
Accountability
5 Interest is defined: Refers to any physical commodity, currency or other tangible goods,
intangibles (such as an equity security, bond, futures contract, swap or option, interest
rates, another index, including indexes that track the performance of a rule-based trading
strategy or the volatility of a financial instrument or another index), any financial instrument
on an Interest, which is intended to be measured by a Benchmark. Depending on the
context, it is assumed that the word Interest also includes the market for such Interest.
6 Methodology is defined: The written rules and procedures according to which
information is collected and the Benchmark is determined.
Islamic
BFI (1)
Islamic
BFI (6)
RBRs
ISI
Islamic
BFI (2)
RBRs
RBRs
RBRs
RBRs
Islamic
BFI (5)
Islamic
BFI (3)
RBRs
Middle East, Africa, and India. The survey found that risk management
in most institutions still tended to operate in isolation rather than being
Islamic
BFI (4)
that the siloed view of risk was often exacerbated by the lack of integration between risk and finance functions (60% of participants disclosed
Islamic
BFI (1)
limited risk-finance interaction), and that [t]his impedes the banks abilW
Western
BFI (1)
Islamic
BFI (2)
RBRs
W
Western
BFI (3)
W
Western
BFI (4)
RBRs
RBRs
RBRs
ISI
Islamic
BFI (4)
RBRs
Islamic
BFI (3)
RBRs
RBRs
RBRs
RBRs
Data quality was also found to be a critical concern, as only around 40%
data which informed their risk models was high (Figure 5).
RBRs
W
Western
BFI (2)
RBRs
W
Western
BFI (5)
Islamic
BFI (5)
RBRs
W
Western
BFI (6)
Islamic
BFI (6)
BFIs may end up rolling out unrealistic risk models and risk management practices, solely to achieve low risk ratings. Consequently it is to be
undertaken and adopted in the spirit of Shariah in order to advance an
ethical and fair approach to the future regional and global development
of Islamic finance. In particular it would seek to advance the secondary
significance. Therefore for the Internal Level, the IFRB framework would
be implemented across each of six Islamic BFIs for a set period, thereby
ness [Ayub (2008)]. From a high level perspective, the Islamic Finance
pendent Supervisory Institution (ISI) that would have legal and opera-
subscriptions made by all participating BFIs, but it would still retain op-
it can first operate at an interim level if need be. Both the first interim
Internal Level (Figure 6) and the final External Level (Figure 7) are
financial quotas by its members). The RBRs would allow all the Islamic
27
relative risk weightings across specific asset classes, and let them see if
Ahmad (1993) argues that Muslim economists have put forth theoretical
BFIs. From one perspective the argument can be made that this Internal
here that the Internal Level framework would only be intended to oper-
ic BFIs carry less risk owing to the fact that they are not based on interest
rates (i.e., less volatility), and that most Islamic finance products are trade
(2010)]. However, at the same time it has also been argued that Islamic
banks are subject to the same risks (credit, market, operational) as West-
ern banks, except that they are magnified because Islamic products are
counterpart BFIs, without first assessing their own risk management ca-
new [Toumi and Viviani (2010)]. Furthermore, Islamic BFIs face unique risks
pabilities and risk levels. In this way the Internal Level framework would
owing to the specific assets and liabilities structure of their balance sheets,
act as a type of primer platform, from which Islamic BFIs could move on
tual breach by Islamic BFIs when managing Profit Sharing Investment Ac-
counts (PSIAs)); (ii) mark-up risk (risk of losses from fixed mark-up rate
used in contract (e.g., Murabaha) owing to a temporal change in relevant
The RBRs would operate at both an Islamic finance product level and
benchmark rate); and (iii) displaced commercial risk (risk of losses which
Islamic BFIs must absorb to ensure PSIAs are paid rate of return equivalent
Islamic product range or overall. Naturally, the higher the number of participating Islamic BFIs, the more accurate the RBRs (in terms of measuring
relative risk) would be. However, the most benefits would be achieved via
an IFRB is the first and necessary step towards empirically proving this
the implementation of the final External Level IFRB framework. This is be-
claim of Islamic superiority right. Otherwise Islamic BFIs might claim their
cause this framework envisages the use of at least six Islamic BFIs and six
Western BFIs, with ideally double that number, i.e., a syndicate of at least
asset-based financing methods, but in reality may fall short of the mark in
terms of the overall risk profile offered. For instance, Muslim consumers
with Islamic windows or established stand alone BFIs. The optimal operat-
ing framework for both levels would be the implementation of public risk
and is asset-based, but which in reality may be far riskier and more prone
tables for a much larger number of Islamic and Western BFIs, as this would
increase the transparency and relative accuracy of risk rankings. The use
it really be said in such instance that this Islamic product complies with
of the ISI as an intermediary between all the participating BFIs would fa-
the spirit of Shariah? The point being made is that Islamic risk bench-
cilitate the acquisition of data needed to determine the RBRs, but without
tained from the BFIs would be protected and not disclosed to any other
participating BFIs, but the ISI would be authorized to monitor and verify
benchmarkings, Islamic BFIs would take one step closer to objective jus-
the data obtained from the BFIs. In principle the External Level framework
systems and practices in Islamic BFIs going forwards; and (iii) afford a new
to Muslim and other consumers. Moreover, even if they are not (per-
it would demonstrate that Islamic BFIs can still work towards reducing
risks for Islamic finance consumers. The underlying Shariah ethos here
28
82
Self assessments
software (44% did not) [Deloitte (2011)]. In addition, many of those used
6 6 6
70
12 6 12
67
16
11 6
17
18
established risk measures such as VaR (Figure 8), scenario analysis and
industry benchmarks (Figure 9) [Deloitte (2011)]. In light of these findings
Scenario analysts
Stress-test
65
53
24
11 12
practices. It is likely to be the case that Islamic BFIs would strive to lower
19
13
19
RBRs to compete with Western BFIs and in doing so, would continue to
Economic metrics
44
Probabilistic analysis
13
38
19
20
19
13
6
13
40
30
37
55
60
80
100
120
Currently in use
catalyst for Islamic BFIs to update and enhance their overall financial risk
Plan to use
Plan to incorporate in next 12 months
No plans to incorporate
42%
37%
32%
NPW/ IRR
21%
11%
Other
11%
money or tied to the maturity or underlying principal sum. For Al-Qaradawi (2001, 54), riba refers to every loan given with pre-conditioned
benefit. The Shariah prohibition therefore expressly condemns the
guaranteed nature of the charge which uses money to make money re-
0%
that the gharar prohibition forbids contracting under conditions of excessive uncertainty and unacceptable levels of risk. Therefore, Western BFI
bank accounts which pay a pre-determined rate of interest on principal
but instead by holistic, equitable and fair risk-sharing initiatives, and the
regardless of risk fall foul of the riba prohibition. However, what about a
ing benchmark risk of the Islamic BFI, which has historical risk rates and
bia, Qatar, UAE), 61% of industry leaders believed risk management re-
seek to emulate Western interest bearing accounts but to all extents and
[Deloitte (2010)]. What is more, 50% of those surveyed did not have a
risk management system to address Islamic financial product require-
ments, and 63.4% agreed that Islamic BFIs were lagging behind in terms
study involving 20 Islamic BFIs across nine countries (with combined as-
sets of more than US$50 billion) 67% of those surveyed used quantita-
tive risk analysis methods (33% did not) and 56% had risk management
29
as forex options which have been used solely for hedging (and not spec-
Stage 1
ulative) purposes. The Waad structure has also been used to arrange
Stage 1 would consist of Islamic and Western BFIs agreeing the blueprint
ing BFIs. In relation to the benchmark design, the IOSCO Principles pro-
gain exposure to Islamic finance products that are not actually owned,
vide at a general level that this should take into account generic design
since the variable leg of the total return swap could be based on the
and Ghoul (2011)]. Also, Shariah futures and options might in the future
eliminate factors that might result in a distortion of the price, rate, index,
they state that relevant factors may include: (i) adequacy of Interest rep-
deferment without actual asset transfer, and used the Islamic product
resentative sample; (ii) size and liquidity of relevant market; (iii) underlying
cise of an option to only within an accepted specific risk range so that the
Firstly, as noted by Allen et al. (2003), in reality all sources of risk (and
volatility) are integrated (e.g., credit risk may drive market fluctuations),
and thus models such as VaR should ideally aim to include known cor-
Stage 2
the accepted types of risk modeling techniques. Thus all external and in-
copula, VcV matrix approach, and Bayesian Markov Chain Monte Carlo
might be 40% (credit risk), 40% (market risk), and 20% (operational risk).
Islamic BFI might comprise 40% (commodity risk), 40% (equity risk), and
20% (liquidity risk), of a 40% overall market risk share. It is submitted that
1. Shariah compliance
Non-existent
1. Shariah risk
2. Prohibition of riba
2. Rate-of-return risk
3. Mark-up benchmark risk
Non-existent
5. Restrictions in requesting
collaterals and penalties
No restrictions imposed
6. Investment accounts
(deposits) are based on a
mudaraba contract
7. Restrictions on secondary
markets and interbank
activities
focusing on methodological issues. This is something that could be developed in time using internal Islamic and Western BFI knowledge of relevant market, credit, and operational risks, and collaboration among IFRB
participants. Instead, the theoretical possibility of developing a comprehensive IFRB framework will be addressed, including the identification of
risks specific to different types of Islamic finance contracts, together with
other relevant operational issues. It is intended that the IFRB framework
30
Islamic banks
Figure 10: Changes to the risk profile caused by the distinct features of
Islamic banks [Salem (2013), 45]
Payment instalments
=
Credit
liquidity risk
Maturity date
Credit risks
Operational
risk
Price + profit
Payment inability
Balance sheet
P and L
fluctuations
Market price
Profit {
Operational risk
Credit risks
Maturity
date
Delayed
payments
Default of payment
Major catastrophe
Earlier leave
PF
Receiving commodity
Variable payment
instalments
Liquidity risk
Market risk
Received back
4
Loss
Fixed payment
Market risk
Market price
fluctuation
1
Commodity price risk
Market risk
Market risk
Figure 11: Credit, operational, and market risks during the lifetime of
Ijarah contracts [Akkizidis and Khandelwal (2008), 71]
Figure 12: Credit, operational, and market risks during the lifetime of a
Murabaha contract [Akkizidis and Khandelwal (2008), 56]
this stage would likely be the most complex as well as controversial stage
of the process. This is because some may argue that Islamic banking risks
levels, the first being the overall bank level where the sources of risk are
banking risks. However, it is opined here that this line of argument is weak,
analyzed (i.e., market, credit, operational risks, and weightings), and the
because although risk profiles may differ, overall risk exposure can be cal-
second level being specific risks arising in each of the main Islamic finance
culated and compared, and this is what matters. The existing financial risk
contracts [Salem (2013)]. For instance, Salem (2013) remarks that market
is very real, and therefore what matters is how precisely such risk can be
risk in Islamic BFIs comprises mark-up, commodity price, forex, and equity
risk, where the first two are specific to Islamic finance contracts, but the
a motorway may have overall risk exposures that differ according to the
last two are identical for conventional banks. It is further noted that opera-
different inherent risk profiles (e.g., age and experience of driver; drivers
appetite for speed; car build and safety level; age of car), but they can
models and the mapping of different risks, resulting in changes to the risk
Stage 3
weightings. For example, it might be agreed that all IFRB BFIs might have
a 20% operational risk weighting, but internal weightings within this 20%
operational risk might differ. This is because Islamic BFI internal risk weight-
ings might instead include Shariah risk, displaced commercial risk, and
and accuracy of the RBRs. In order to maximize the accuracy of the IFRB
mark-up risk. This would require an ex ante analysis of risk typologies for
it is envisaged that there would be a six month test phase during which
each Islamic finance contract offered by relevant Islamic BFIs. For example,
the accuracy of the IFRB could be tested prior to the final official rolling
Ijara contracts involve leasing an asset (commodity price risk) together with
out phase. Consequently, it is suggested that during the first six month
commodity price risk arises (asset acts as collateral) but market risk arises
in the shape of price risk (i.e., mark-up risk) (Figure 12) [Salem (2013)]. The
lar fashion to the way credit rating agencies such as Fitch, Moodys,
finance products, and for all Islamic and Western BFIs to agree an equiva-
and Standard & Poors work. In theory the historical data obtained by
the ISI could also help to improve the accuracy of financial risk modeling
within Islamic BFIs, although disclosure of such information would ne-
the rolling out of the risk benchmarking process, but at the same time it is
31
2
P and L
fluctuations
Market risk
Market risk
Level of investment
Share price fluctuations
Market price
Last equity
payment and/or
sum of payments
Loss
Payment instalments
Profit
Operational risk
Credit risk
Liquidity risk
Credit risk
Liquidity risk
Payment inability
Major loss
Inability to carry on
business
Operational risk
Business risk
Balance sheet
fluctuations
End of partnership
Loss
Loss
Investment principle
Profit
1
2
opined that the process is realistic and can be achieved. In fact there are
two main difficulties that may arise in this type of risk modeling. The first
ate Islamic finance risk models could be developed and applied. Moreover,
it also means that relative risk weightings and a risk hierarchy could also
change over time, and static risk measurements would fail to capture
ing instruments such as the Mudaraba pose the highest risk, followed by
contract, risk weightings for credit, operational, market, and liquidity risks
goods deferred delivery contract), and then lowest risk Ijara and Mura-
or diminishing (Figure 14) in nature. The second difficulty is that the role
baha (Figure 15 and Figure 16) (a more fixed and predictable margin). Con-
more than capable of developing and applying these types of risk model-
ing methodologies. For instance, BFIs full historical loss data for specific
BFI, as well as the strength and robustness of the BFIs existing risk man-
for defining stop loss sale value, or in static and dynamic analysis for esti-
mating current and future VaR for market risk in Musharakah, Mudarabah,
liquidity risk measurements systems in place for all participating BFIs. The
importance of such systems would lie, not in disclosing the proprietary
32
Nevertheless, the fact that the functionality of the different types of Islamic
P and L
fluctuations
Market risk
Level of investment
Last equity
payment and/or
sum of payments
Inability to carry on
business investment
Investment principle
Liquidity risk
Credit risk
Liquidity risk
2
1
Balance sheet
fluctuations
Inability to carry on
business
Operational risk
Business risk
Major loss
Operational risk
Loss
Cycle of investments
Major loss
Loss
Business risk
a manner supervised and approved by the ISI in accordance with the es-
sion and uniformity, and owing to its somewhat restrictive nature it has
At the same time, arguments have been made that owing to the ethical
and equitable framework of Islamic finance, Islamic finance contracts are
Conclusion
safer than Western counterparts (in terms of financial risk), since on the
whole they are finance contracts that are based on identifiable underly-
ing assets. The limited financial losses incurred by Islamic BFIs during
lamic faith. In practice this means that Islamic BFIs must provide finan-
the recent financial crisis would seemingly testify to such a notion. Nev-
Nakheel sukuk in 2008-2009 owing to the Dubai property crash, also bear
testament to the fact that Islamic finance is not immune to large financial
losses. In light of these propositions it has been argued here that the de-
lic and private Islamic and Western BFIs would contribute to significantly
risk management systems and practices in Islamic BFIs, and it could pro-
cluding, but not limited to, the Murabaha, the Mudaraba, the Musharaka,
Islamic finance products. It has been seen that the External Level IFRB
and the ijara contracts, as well as Islamic sukuk. Nevertheless, while Is-
framework would map out all the most relevant market, credit, opera-
tional, and compliance or Shariah risks. It would also utilize the IOSCO
still undergoing further development. For example, it has been seen that
operational RBRs for all participating BFIs. From one perspective it could
33
enable continuous risk management improvement, and would lend transparency to Islamic finance operations. It might also be argued that the
External Level IFRB framework could galvanize participating Islamic BFIs
into action vis--vis risk management capabilities, and would also significantly contribute to overcoming the paradigm blindness regarding the
need for competitive advantage at all costs. Consequently in light of the
underlying ethos and principles of Shariah, it is submitted that the collaborative benchmarking of risk of Islamic and Western BFIs could lead
to significant new benefits for all participating BFIs across-the-board.
References
Abdulkader, T., S. Cox and B. Kraty, 2005, Structuring Islamic Finance Transactions, London
Euromoney Institutional Investor PLC
Ahmad, A., 1993, Contemporary Practices of Islamic Financing Techniques, Jeddah: Islamic
Research and Training Institute, Islamic Development Bank
Akkizidis, I. and S. K. Khandelwal, 2008, Financial Risk Management for Islamic Banking and
Finance, Hampshire: Palgrave Macmillan
Allen, L., J. Boudoukh, and A. Saunders, 2003, Understanding Market, Credit and Operational
Risk: The Value at Risk Approach, Oxford: Blackwell Publishing
Allen, S. L., 2013, Financial Risk Management: A Practitioners Guide to Managing Market and
Credit Risk, Chichester: John Wiley & Sons
Al-Qaradawi, 2001, Banks Interest is the Prohibited Riba, Cairo: Wahba Publishers
Anand, G. and R. Kodali, 2008, Benchmarking the benchmarking models, Benchmarking:
An International Journal, 15(3): 257-291
Arora, N., J. R. Bohn, and F. Zhu, 2005, Reduced Form vs. Structural Models of Credit Risk:
A Case Study of Three Models, Moodys KMV Company: 1-39
Attalah, C. C. and W. A. Ghoul, 2011, The Wad-Based Total Return Swap: Sharia Compliant or
Not?, The Journal of Derivatives, 19(2): 71-89
Ayub, M., 2008, Understanding Islamic Finance, Chichester: John Wiley & Sons
Bank for International Settlements (BIS), 2006, Basel II: International Convergence of Capital
Measurement and Capital Standards: A Revised Framework - Comprehensive Version, Bank
for International Settlements (June)
Black, F. and M. Scholes, 1973, The Pricing of Options and Corporate Liabilities, Journal of
Political Economy, 81: 673-654
Buehler, K., A. Freeman and R. Hulme, 2008, Owning the Right Risks, Harvard Business
Review (September) 1-9
Camp, R. C., 1989, Benchmarking: The Search for Industry Best Practices that Lead to
Superior Performance, Milwaukee: ASQC Quality Press
Chinoy, S, 1995, Interest-free banking: the legal aspects of Islamic financial institutions,
Journal of International Banking Law, 10(12): 517-524
Deloitte, 2010, The Deloitte Islamic Finance leaders survey in the Middle East Benchmarking
practices, Biannual Survey (Issue 1)
Deloitte, 2011, Empowering Risk Intelligence in Islamic Finance, Managing risk in uncertain
times, Deloitte in the Middle East
DeLorenzo, S.Y.T., 2007, Symposium: Islamic Business and Commercial Law: Shariah
Compliance Risk, Chicago Journal of International Law, 7(2): 397-408
Delpachitra, S. and D. Beal, 2002, Process benchmarking: an application to lending products,
Benchmarking: An International Journal, 9(4): 409-420
Ernst & Young, 2012, Global Islamic assets are expected to reach $1.8 trillion by 2013 (10
December, Dubai), available at: http://www.ey.com/EM/en/Newsroom/News-releases/10-Dec2012Global-Islamic-assets-are-expected-to-reach---1-8-trillion-by-2013
Fadeel, A., 2002, Legal aspects of Islamic finance, in Karim, R. A. A. and S. Archer (eds), Islamic
Finance: Innovation and Growth, London: Euromoney Books
Fong, S. W., Cheng, E. W. L., and Ho, D. C. K., 1998, Benchmarking: a general reading for
management practitioners, Management Decision, 36(6): 407-418
Gerken, A., N. Hoffmann, A. Kremer, U. Stegemann, and G. Vigo, 2010, Getting risk ownership
right McKinsey Working Papers on Risk, 23: 1-13
Hanif, A., 2008, Islamic finance an overview, International Energy Law Review, 1: 9-15
Hull, J. C. and A. White, 2000, Valuing Credit Default Swaps I: No Counterparty Default Risk,
(April) NYU Working Paper No. FIN-00-021
Ioannis, A., K. Lampros, Z. Ioanna and B. Vivianne, 2006, Integrating Market, Credit and
Operational Risk: A Complete Guide for Bankers and Risk Professionals, London: Risk Books
34
High-level Debate
Credit Valuation
Adjustment: The
Devil is in the Detail
Sylvain Prado Consultant, Capco
Bhavdeep Virdee Consultant, Capco
Abstract
On September 15, 2008 Lehman Brothers unexpectedly
the USA at that time. Lehman could no longer honor its debt
are often vague, confusing and over technical. With this pa-
with the bank were facing huge counterparty risk with the
Section 5 concludes.
35
of the regulatory capital framework for the trading book. Banks need a
CVA risk capital charge to comply with the latest Basel III requirements
the risk of changes in the mark to market (MtM) values of over the counter
[BIS (2011)]: In addition to the default risk capital requirements for coun-
occurred during the financial crisis arose from MtM movements rather
ratings-based (IRB) approaches for credit risk, a bank must add a capital
This is made up of the default free value, the debit valuation adjustment
tive product. An example with an interest rate swap describes below the
components of CVA.
can apply Monte Carlo simulation to generate various scenarios for the
floating interest rate. The estimation of the Expected Exposure (EE)3, de-
CVA enables the bank to decide on the best way to manage exposure and to choose the appropriate hedging strategy.
Figure 2 illustrates the structure of the cash flow for an interest rate swap.
Figure 3 illustrates the steps to compute CVA for the interest rate swap.
tions, however, need to be answered: At what price should the protection be bought? How much credit protection should be bought? What is
the right time to re-hedge the underlying counterparty risk? Can we link
credit and market risks and then hedge the combined risk? How does a
trader manage wrong way risk or the correlation between the exposure
and default probability of an obligor?
These questions should be addressed by the bank and take into account
36
Credit administration
Counterparty credit
exposure and limit
management
Accounting
Accountants need to add
unrealised losses on books
Impact of risky
MtM value on?
Funding
Traders need to source
additional funds to address
the liquidity deficit
Counterparty A
Counterparty B
Receive floating
@ LIBOR + Y.YY%
with quarterly payments
Liquidity risk
Manage and monitor the
funding gap
Market data
acquisition
Calibration to initial
yield curve
Boot strapping
Calibrated model
parameter
values are
the inputs
Monte Carlo
simulation to
generate spot yields
based on zero
coupon yield
0.06
0.05
Model calibration
LIBOR
Market
yieldRate
curve
Parameters
calculated
using
various
methods like
Maximum
Likelihood
Estimation
and Kalman
filtering
0.04
0.03
0.02
0.01
0
100
50
0
100
80
60
40
20
Forward curve0
-3
x 10
3.5
Simulation to
generate
exposure profiles
2.5
Expected Exposure
Two methodologies to
generate EPE values
Scenario generation
Simulations
1.5
0.5
0
15
10
100
60
70
80
90
50
30
40
20
10
Periods
Simulations
Wron
g
way
of sim risk impa
c
ulate
d inte t through
c
rest r
ate to orrelation
PD
CVA distribution
Payer or receiver
perspective
Figure 3 The steps to compute CVA for the interest rate swap
2.5
Population in Buckets
Estimation of the
CVA and
adjustment of a
new Mark to
Market (MtM)
1.5
0.5
0
2.4
2.6
2.8
3.2
3.4
3.6
3.8
4.2
-4
CVA in % of the
nominal amount
x 10
37
As specified in Basel III, there are three possible combinations in the cal-
Documentation
Adequacy of documentation of counterparty risk management system and process
culation of the Total Counterparty Credit Risk Capital (CCR) charge. And
the formulas developed by regulators are not really intuitive.
For banks with Internal model method (IMM) approval and market-risk
Approval process for risk pricing models and valuation systems used by front and back-
total capital charge is the sum of the advanced CVA risk capital charge
office personnel
Validation of any significant change in counterparty risk measurement process
and the higher value of either the IMM capital charge based on current
Data acquisition
Integrity of management information system
S i-1 .ti-1
S i .ti
-exp LGD MKT
LGD MKT
For banks with IMM approval and without specific-risk VaR approval
Table 1
for bonds, the Total CCR Capital is the sum of the standardized CVA
CVA incorporates portfolio dynamics.
risk capital charge and the higher value between the IMM capital charge
based on current parameter calibrations for EAD and the IMM capital
reduce the total exposure. A margin agreement, also established contractually, requires that a counterparty posts additional collateral when
2
hedge
K=2.33. h.
the exposure exceeds a threshold. When the exposure goes below the
hedge
Bi
h = 1 (one-year horizon)
Rating
AAA
0.7%
AA
0.7%
0.8%
BBB
1.0%
2.0%
CVA for capital measurement is not the same as CVA for pricing.
The CVA capital charge depends on the capital methodology approv-
B
CCC
Weight wi
3.0%
18.0%
Netting
For all other banks, it is the sum of the standardized CVA risk capital
charge and the standardized method based capital charge or the sum
Trade level
CVAs
OTC
Portfolio
Collateral
38
file.
The previous section outlines the steps to calculate CVA from a pric-
quired to clearly understand the linkages between the various components of CVA.
ence data across the front office, risk, and finance in a harmonized man-
risk and ensures compliance with the latest regulations. On-going vali-
narios and market risk factors across all products. Systems infrastructure
product taxonomy.
Credit spread
calibration for CVA
Collateral close-out
risk
Collateral close-out
contagion
Correlation risk
CVA hedging
reveal that a previously profitable deal may not be so under the new regime. Banks will learn how new levers can be pulled to influence pricing
and capital consumption, thereby achieving the desired balance between
risk and reward.
Conclusion
This article gives the reader an opportunity to clarify, beyond usual theoretical literature, the key considerations for a bank when developing a
CVA framework.
CVA is used for the pricing adjustment of OTC derivatives and as an additional component of reserves calculations. As a consequence, the deal
assessments, the hedging techniques and other strategic decisions of a
bank are strongly impacted.
CVA contains, by definition, a complex blend of credit and market risk
parameters. Given the rapid changes in the regulatory landscape and the
need to be competitive, it is crucial for an organization to create a flexible and robust computational finance workflow that effectively links data
management, statistical modeling, technology, and decision-making.
5 It also needs to be consistent for all asset classes (rates, FX, equity, commodities, credit).
39
References
40
Bank of International Settlements (BIS), 2011, Basel III: A global regulatory framework for more
resilient banks and banking systems, Basel
Brigo, D. A. Dalessandro, M. Neugebauer, and F. Triki, 2007, A Stochastic Processes Toolkit for
Risk Management
Dodd Frank Act, 2010, Public Law 111 - 203 - Dodd-Frank Wall Street Reform and Consumer
Protection Act, U.S Government Printing Office
Office of the Comptroller of the Currency (OCC), 2011, Fed OCC guidelines handbook, U.S.
Department of Treasury
Zhu, S. H. and M. Pykhtin, 2007, A Guide to Modeling Counterparty Credit Risk, GARP Risk
Review, July/August 2007, available at SSRN: http://ssrn.com/abstract=1032522
High-level Debate
Aligning Marketing
and Finance with
Accepted Standards
for Valuing Brands
James Gregory Founder and CEO of CoreBrand
Michael Moore Professor in Residence of Accounting at the Loyola Marymount University
Abstract
With this paper we hope to create awareness and support
among the financial community for the need to have consistent brand measurement and metrics that tie investments
in corporate and product brands to financial value. There
are many obstacles to overcome before the financial community accepts such brand value measurements. Our initial
proposal suggests that brand value created (or lost) by the
investment in brands be reported in the Management Discussion and Analysis section of the financial reports. This will
promote a consistent methodology for valuing brands for internal management purposes and will provide to all external
parties full disclosure about the benefits and value created
(or lost) by investment in brands.
41
Introduction
and the International Accounting Standards Board (IASB) have been very
locating resources for marketing activities of the firm with little guidance
as to the values created by these expenditures.
Investors and analysts attempt to value and compare firms and try to
predict how assets owned by the firms will produce future income with
bination are measured and reported in the financial records under both
Marketing assets
With respect to branding and other marketing intangibles there are no-
financial information.
books at adjusted historical cost and are not written up to market. Like-
wise, purchased intangibles are recorded at cost and are adjusted down
by amortization or impairment and never written up to market.
The current debate about whether or not intangible assets such as the
vestment and valuation standards, using metrics that are simple, trans-
gap between those who espouse such treatment and the policy makers
at FASB and IASB. Recording assets at fair value using Generally Ac-
comes, and to financial returns both short term and over time.
not adequately reflect the fair value of marketing assets or in most cases
does not even reflect the existence of an asset. Most feel that such a
The Improving Financial Reporting (IFR) Project has as its goal to facili-
such that financial returns from corporations will be driven and measured
by buyer behavior in markets over time and to ensure that marketing is
at the table when reporting of brand value is required for internally de-
accounting may not occur for several years. Development of a model for
veloped brands.
42
focuses on the price that would be received upon the sale of an asset or
paid to transfer a liability. It is an exit price rather than a price that would
observable inputs are not available. This allows for situations where
there is little or no market activity for the asset or liability at the mea-
Valuation issues
allow for situations in which there is little, if any, market activity for the
For both FASB and IAS, valuation techniques that are consistent with
the market approach, income approach, and/or cost approach are used
to measure fair value. A market approach uses prices and other relevant
In addition to the general fair value models from FASB and IAS and specific
Interbrand, and Millward Brown, other models have been proposed, nota-
market approach might be one that uses market multiples derived from a
International Valuation Standards (IVS) -2011. This study sets standards for
els, and the multi-period excess earnings method. The cost approach
is based on the amount that currently would be required to replace the
In order to persuade the FASB and the IAS that marketing assets should
There are many inputs (assumptions) that market participants use in pric-
a fair value model. This project involves creation of general principles and
a valuation technique and/or the risks inherent in the inputs to the valuation
entity. Unobservable inputs are inputs that reflect the reporting entitys own
will have several levels of impact metrics: customer level, market level,
Level 1 inputs are quoted prices in active markets for identical assets
or liabilities that the reporting entity has the ability to access at the
equity securities as an example, debt securities that a firm has the intent
measurement date.
Level 2 inputs are inputs other than quoted prices included within
Level 1 that are observable for the asset or liability, either directly or
cost approach). Debt and equity securities that are purchased principally
to sell in the near term are classified as trading securities and reported
gains and losses are included in earnings. Debt and equity securities not
classified as above are classified as available for sale securities and re-
Inputs other than quoted prices that are observable for the asset or
Problems abound even with fair value balance sheet measurements that
43
MD&A section of the annual report will have a profound impact on the
2010 among clients of the big four accounting firms that were related
practice for external users. This single change will make marketing more
Marketing assets do not fit well in this current framework of fair value
measurement for inclusion of such assets on the balance sheet and many
marketing were wise and provided the company with an effective return.
Marketing activities can be highly efficient for value creation, but not
For reasons stated above, rather than entering the fair value of marketing
assets on the balance sheet, MASB is proposing that disclosure be made
assets have grown in value there is less understanding than ever of what
drives that value. There is no doubt that accounting standards for intan-
of management.
gible assets will eventually be changing worldwide. CFOs who see their
role from the accounting perspective will wait until IASB standards for
Examples:
future. But from the financial leadership perspective, CFOs will embrace
and encourage forward thinking ideas about what drives the value of intangible assets. Those CFOs who lead the charge will be providing better
and 20% over the past three years. We estimate this value using the
Valuator.
Conclusion
The following note is an example of a proposed MD&A disclosure for
and will serve as the primary forward looking marketing KPI in Corporate
2.7% from a year ago, and down 3.8% over the past three years. We
estimate this brand value using the methodology provided by BrandTop, LLC,a MASB qualified Brand Valuator.
We seek to have CFOs, CPAs, and the world of finance join MASB to
achieve consistent, comparable, credible and actionable brand valua-
44
tions for both externally and internally developed corporate and product
goal is to tie reliably marketing actions to customer impact, market outcomes, and financial returns both short term and over time.
For further information please visit: http://www.themasb.org
References
Chasan, E., 2012, Asset Valuations Trip Up Audits, Wall Street Journal, 22 May
Financial Accounting Standards Board (FASB), 2006. Fair Value Measurements, Statement of
Financial Accounting Standards No. 157, available at: www.fasb.org
45
46
High-level Debate
CCP Regulations:
Posing Challenges
for Both CCPs and
Regulators
Jennifer Liu Principal Consultant, Capco
Abstract
Given the focus on central counterparty (CCP) clearing in
the post-financial crisis world, it is no surprise that clearing
houses today face more scrutiny from regulators and central banks than ever before. However, given the complicated
legal entity and organizational structure of CCPs, regulating
and providing oversight of clearing corporations can be challenging. This is because different subsidiaries tend to fall into
different jurisdictions and are subject to different regulatory
bodies. In addition, in light of the different products cleared,
CCPs can be subject to product-specific regulations. For
clearing houses, the complexity of complying with different
regulations across the world can be taxing. For regulators,
the intricate structure of global clearing houses poses challenges when applying appropriate oversight.
47
Introduction
Background
In the wake of the financial crisis and in response to the G20 commit-
CCPs have long been used by derivatives exchanges and a few securities
exchanges. In recent years, they have been introduced into more securi-
ties markets, including cash markets and OTC fixed income markets, be-
instrument used to provide funding for dealers and banks, can now be
cleared through several clearing houses in Europe, the US, and Canada.
clearing services, can be traced back to more than 100 years ago when
transactions are cleared by a centralized counterparty. Through a process called novation, the CCP becomes the seller to the buyer and the
buyer to the seller. A clearing house usually provides the CCP service
LCH was merged in 2003 with Clearnet SA, which was formed in 1969
and manages the counterparty default risk by leveraging its built-in risk
Clearing houses
Location
ing members net obligations to post margins and settle contracts, the
LCH.Clearnet
UK
Europe
US
Eurex Clearing
Europe
UK
Derivatives
Equities
Bonds
Repo
Secured funding
Securities financing
Energy transactions
CME Clearing
US
Europe (through
CME Europe)
ICE Clearing
US
Europe
Canada
role of the CCP has expanded over time. Since the financial crisis, the
importance of CCPs in providing financial stability through their risk management arrangements cannot be overstated.
48
Holding company
2009 to provide the risk management framework, operational processing, and clearing infrastructure for ICE Clear Credit and ICE Clear Europe.
CDCC (Canadian Derivatives Clearing Corporation) is the product of a
merger between Montral and Toronto options clearing houses in 1977
and recently launched a fixed-income CCP service in addition to its derivatives clearing business.
Operating entities
LCH.Clearnet LLC
Location: New York
Organized in Delaware
Regulated as a
Derivatives Clearing
Organization (DCO)
by the US Commodity
Futures Trading
Commission (CFTC).
LCH.Clearnet Limited
Location: London
Incorporated in UK
LCH.Clearnet SA
Location: Paris
Incorporated in France
Regulated as a
Recognized Clearing
House by the Bank of
England, as a Derivatives
Clearing Organization
(DCO) by the US
Commodity Futures
Trading Commission
(CFTC).
Regulated as a Credit
Institution and Clearing
House by a regulatory
college consisting of the
market regulators and
central banks from the
jurisdictions of France,
Netherlands, Belgium and
Portugal.
Regulated as a
Recognized Overseas
Clearing House by the
Bank of England and by
other market regulators
and central banks in
jurisdictions in which
business is carried out.
Pending Derivatives
Clearing Organization
(DCO) application with
the CFTC.
Regulated by other
market regulators
and central banks in
jurisdictions in which
business is carried out.
Payment systems are
overseen by the Bank of
England.
regulating and providing oversight of clearing corporations can be challenging as different subsidiaries tend to fall into different jurisdictions and
are subject to different regulatory bodies. In addition, given the different
products cleared, CCPs can be subject to product-specific regulations.
Regulation by location
From a regional perspective, in Europe clearing houses with entities located or business conducted in the UK are regulated by the Financial Services
Luxembourg
subsidiary
Source: LCH.Clearnet
subject to their local regulators. For example, Eurex Clearing is a company incorporated in Germany and licensed as a credit institution under
In the US, clearing houses such as CME Clearing and FICC are registered
tiel (ACP) (France). Its operating entities are regulated by CFTC in the US,
Regulation by products
for illustration.
depending on their product offerings. For example, the US Commodseeks to provide clearing services with respect to futures contracts, op-
After the crisis, regulators in the world worked in tandem to review the
existing financial infrastructure and drafted new regulations that aim to in-
ment not only affects clearing houses in the US but also in the rest of the
with the CFTC now, including CME Clearing US, Eurex Clearing, LCH.
ity Futures Trading Commission (CFTC) requires any clearing house that
49
more robust and frequent stress tests, and processes for orderly resolution
fees, risk model data, aggregate transaction volumes, rules and pro-
Impact for CCPs: The increasing amount of information and data re-
require the CCP to build additional systems to meet the requests. The
client account structures for both assets and positions across all
asset classes. Portability must also be offered and possible for each
account structure.
enforced.
management (BCM) means that CCPs may need to have more re-
data depositories can pose a challenge for CCPs to meet all the re-
from the need for resources and technology development, as well as the
quirements in time.
means CCPs need to report the transactions as required and may also
calculate the minimum size of its default fund. Also, CCPs are now
means that technology will play a critical role for CCPs aiming to meet
the rules are still evolving so timelines, priorities, and scope will be
50
Impact for CCPs: The capital required of CCPs comes at a cost. While
On the other side of the coin, regulators are facing their own challenges
CCPs try to meet the capital requirement, they also need to under-
clearing facilities.
capital costs and operating costs. This includes the cost of technol-
ahead of the game, CCPs need to look internally and externally for op-
CCPs are able to pass the cost to their clearing members. However,
the market:
tegic roadmap and carefully plan out and prioritize the technology
tives space given the demand from banks to meet the central clearing
the interim, CCPs may also consider forming alliances with consulting
ing house can offer to its clients, the more capital efficiency its clients
tunities.
resources. They will then need to formulate strategies that help opti-
tions, and we can expect to see even more activities in this area. For
example, ICE and NYSE announced their merger in December 2012.
Also, EuroCCP, the European unit of the DTCC (Depository Trust &
practices.
51
Current state
Jurisdiction A
Authority
X
Authority
Y
Jurisdiction B
Authority
Z
Authority
X
Authority
Y
Subsidiary 1
Jurisdiction C
Authority
Z
Authority
X
Subsidiary 2
Authority
Y
Authority
Z
Subsidiary 3
CCP
Regulation within jurisdiction
Regulation across jurisdiction
Future state
Jurisdiction A
Authority
X
Authority
Y
Jurisdiction B
Authority
Z
Authority
X
Subsidiary 1
Authority
Y
Jurisdiction C
Authority
Z
Subsidiary 2
Authority
X
Authority
Y
Authority
Z
Subsidiary 3
CCP
52
CCPs, not only within the same jurisdictions but cross-border and
globally. For example, the Dodd-Frank Act provides that the CFTC will
down and bottom-up approaches are adopted when rules are pro-
the CFTC and SEC will regulate mixed swaps. This requirement
References
Understand the broader impact in the whole market When considering new rules, regulators need to carefully evaluate the potential
mentioned earlier, the increasing cost of central clearing may curb the
Summary
Regulation of CCPs is a critical task in todays global market. CCPs have
CFTC Office of Public Affairs, 2012, Q & A Proposed Rules and Interpretive Guidance Further
Defining Swap, Security-Based Swap, and Security-Based Swap Agreement; Regarding
Mixed Swaps; and, Governing Books and Records for Security Based Swap Agreements,
Commodity Futures Trading Commission
Commodity Futures Trading Commission (CFTC), 2011, Derivatives Clearing Organization
General Provisions and Core Principles, Federal Register, 76, 216
CPSS-IOSCO, 2012, Principles for Financial Market Infrastructures, Bank of International
Settlement and International Organization of Securities Commissions
DoddFrank Wall Street Reform and Consumer Protection Act, 2010
The Eurofi, 2012, Improving Collateral Management Efficiency in the Context of OTC Derivative
Reforms, The Eurofi Financial Forum, September
The European Commission, 2013, Regulatory Technical Standards on Capital Requirements
for Central Counterparties, Official Journal of the European Union, February
The European Commission, 2013, Regulatory Technical Standards on Requirements for
Central Counterparties, Official Journal of the European Union, February
The European Commission, 2013, Implementing Technical Standards on Requirements for
Central Counterparties, Official Journal of the European Union, February
Financial Stability Board (FSB), 2012, Strengthening Oversight and Regulation of Shadow
Banking, An Integrated Overview of Policy Recommendations, Consultative Document
become an important component of the global financial market infrastructure. Unless properly managed and well capitalized, the CCPs can
potentially pose significant systemic risk to the market, as they take on
additional concentrated risk. There have been heated discussions about
the vulnerability of CCPs in distressed market situations. Some even
start to apply the term too big to fail to CCPs. More standardization
and transparency of CCPs can therefore be anticipated as the market
evolves. In the meantime, regulators need to have a holistic view of the
market infrastructure to effectively develop and enforce regulations that
ensures the safety of the market while encouraging the healthy growth of
market. Like the old saying Rome was not built in a day, we can expect
this to be a long, albeit exciting, journey.
53
54
High-level Debate
The US Home
Foreclosures and
Federal Funds Rate:
Is the Channel
Effect Muted?
Gke Soydemir College of Business Administration, California State University, Stanislaus
Andres Bello College of Business Administration, University of Texas-Pan American
Tzu-Man Huang College of Business Administration, California State University, Stanislaus
Abstract
We provide empirical evidence on the postulation that the
channel effect in the US is muted. The estimation results
from a multivariate model reveal that the movements in the
federal funds rate do not appear to have the intended timely
and pronounced impact on foreclosures starts and purchase
prices. The response of the home mortgage interest rates
to the federal funds rate however, is significant and positive
but the effect is not sustained. These findings are consistent
with the view the channel effect in the US is relatively muted.
55
Introduction
Theoretical background
In the summer of 2007, the US stock market crashed and brought with it
the housing market meltdown. The meltdown spread beyond the hous-
ing market and became global resulting in a systemic crisis in 2008.1 The
economy. The nature and slope of the aggregate supply schedule can
Federal Reserve responded to avert the crisis and minimize its impact in
the US by initially utilizing traditional tools such as the federal funds rate.
Changes in the federal funds rate did not have an immediate impact on
Alan Greenspan noted in the recent past that the Federal Reserve be-
For example, the Keynesian theory [Keynes (1936)] posits that under the
mortgage rates when the latter failed to respond as expected to the Fed
rates had become gradually decoupled from monetary policy even ear-
lier in the wake of the emergence, beginning around the turn of this
ments.
The basic textbook Classical economics asserts that prices are comFurther, Rubio (2008) in investigating the postulated relationship between
pletely flexible (both in the short run and long run) and with a relative-
fixed and variable rate mortgages, business cycles, and monetary policy,
quotes Ben Bernanke: [...] the structure of mortgage contracts may mat-
ter for consumption behavior. In countries like the United Kingdom, for
librium, only have an effect on the price level with no significant impact
on the level of output. Thus in the Classical model, the levels of output
cash flows. [...] In an economy where most mortgages carry fixed rates,
such as the United States, that channel of effect may be more muted. I do
not think we know at this point whether, in the case of households, these
Utilizing two decades of data, the aim of this paper is to conduct empirical
Fiscal policy implications of the Classical model on the level of output are
tests on the postulation that the federal funds rates effect on US home
funds. The resulting rise in the federal funds rate causes a decline in invest-
activity and thus shed some light on an issue that has yet to be studied
ian economics, the crowding out effect is partial, and in the case of the
The estimation results from a multivariate model indicate that the recent
movements in the federal funds rate do not appear to have intended
timely and pronounced impact on foreclosures. Further, the housing prices fail to respond to decreases in the federal funds rates. The findings are
consistent with the view that the correction in the housing market is likely
to continue as a policy independent process until all excess inventories
56
1 Financial Times, 2008, Credit turmoil flagged up as risk to global outlook, January 10
2 Greenspan, A., 2009, The Fed Did Not Cause the Housing Bubble, Wall Street Journal,
Opinion Column, March 11
3 Bernanke, B., 2007, The Financial Accelerator and the Credit Channel, Remarks at a
Conference at the Federal Reserve Bank of Atlanta.
4 In the extreme case of a horizontal aggregate supply curve, the price level remains constant
and aggregate demand management becomes most effective in securing full-employment.
policy stance of the Federal Reserve. Indeed the extent to which the fed-
scribes the data and econometric methodology. Section four reports em-
Y=C+I+G
(1)
The sample interval spans the third quarter of 1980 through the fourth
Y = C + S + T
(2)
Datastream. The Fed policy proxy variable used in this study is the effective federal funds rate (FEDFUND). Foreclosure starts (FORCST) are for
all residential mortgage loans, as are the home purchase prices (PRICE).
Home mortgage rates are for the 30 year fixed rates (HOMIRATE). Re-
savings;
gional variables considered in the study are the foreclosure starts from
S = 0 - 1F
(3)
the regions of the West (WEST), South (SOUTH), Northeast (NORTHEAST), and North Central (NORTHCENTRAL). Price series were not avail-
Figure 1 plots the foreclosure starts from the third quarter of 1980 to the
fourth quarter of 2012. It appears that the jump in foreclosures starts in
By substitution and solving for F, the reduced form can be expressed as;
the fourth quarter of 2006 with a spike from .47 in the third quarter to
.57 in the fourth quarter, and then exhibiting an exponential rise. A visual
inspection of the graph shows that there may be evidence of a slight
1. When the null hypothesis holds, (1/ 1) = 0, interest rates (in this case
until the third quarter of 2006 both series long-run trend appears to be
either somewhat flat or slightly upward for foreclosure starts and down-
ward for the Federal Funds rate consistent with either no presence of an
association between the two or a weak negative association. However,
Further, apart from the two schools of thought, there is some evidence in
toward the latter part of the sample, after the third quarter of 2006, the
[Fraser et al. (2008)], a result in favor of the view that much of the price
coefficient of -.93 in this sub-sample period and -.58 in the entire sample. These findings are consistent with the view that during much of the
ior [Shiller (1984), Kyle (1985), DeLong et al. (1990), Daniel et al. (1998),
Barberis et al. (1998), Hong and Stein (1999), Lui et al. (1999)]. Driven by
price alone, momentum occurs when agents buy after price increases
and falling of the foreclosure rates. These findings are also consistent for
and sell after price decreases. In this scenario agents expect a price rise
the variables in the form of logarithmic first differences which would cap-
ture any leveling off effect or decrease in the rate of increase. To measure
causes demand to fall even further. An indirect result of this is the impact
on foreclosures. Continuing decrease in demand means there are now far
57
Number of
Cointegrating
equations
Value
1,6
1,4
None
1,2
1
0,8
Eigenvalue
Trace
P-value**
Maximum
Eigenvalue
P-value**
0.254
44.951
0.091
26.335
0.072
At most 1
0.111
18.616
0.521
10.598
0.687
At most 2
0.082
8.018
0.464
7.724
0.408
At most 3
0.003
0.294
0.588
0.294
0.588
0,6
0,4
0,2
Quarters
Unit root tests are employed to analyze the time series properties of the
data to avoid the possibility of finding spurious relationships. In particular, the Augmented Dickey-Fuller (ADF) test [Dickey and Fuller (1979,
1981), Enders (1995)] fails to reject the null hypothesis of non-stationarity
in logarithmic levels for most of the series, but we reject the null for the
Value
1,6
log returns for all the variables. Given that most of the series are non-
1,4
stationary in levels, differences of the series are taken. The AIC and SBC
1,2
criteria suggest that the appropriate number of lags to be used is two and
0,8
sample and to avoid any issues with over parameterization and thus the
0,6
0,4
Q3 1992
Q1 1993
Q3 1993
Q1 1994
Q3 1994
Q1 1995
Q3 1995
Q1 1996
Q3 1996
Q1 1997
Q3 1997
Q1 1998
Q3 1998
Q1 1999
Q3 1999
Q1 2000
Q3 2000
Q1 2001
Q3 2001
Q1 2002
Q3 2002
Q1 2003
Q3 2003
Q1 2004
Q3 2004
Q1 2005
Q3 2005
Q1 2006
Q3 2006
Q1 2007
Q3 2007
Q1 2008
Q3 2008
Q1 2009
Q3 2009
Q1 2010
Q3 2010
Q1 2011
Q3 2011
Q1 2012
0,2
Foreclosure starts
To test the relationship with one lag, we use a VAR model. Impulse re-
sponse functions (IRFs) are generated from the VAR model to analyze
it does not impose a priori restrictions on the structure of the system and
In particular, the IRFs trace the response of one variable to a one stan-
standard errors or confidence intervals is equivalent to reporting regression coefficient without t-statistics. For statistical inference purposes, we
Z(t) = C +
use confidence bands around the mean responses based on Doan and
Littermans (1986) Monte Carlo simulation technique. Accordingly, when
A(s)Z(t
s) + e(t) (5)
s=1
the upper and lower bands carry the same sign, the responses are considered to be statistically significant at the 95% confidence level.
efficients, m is the lag length and e(t) is a vector of random error terms.
Lastly, Runkle (1987) states that to avoid the problem of over parameterResponse of LOG(HOMIRATE) to LOG(FEDFUND)
ization average situations are models with nine parameters per equation
.04
for 40 data points, 30 for 120 and 70 for 400. The model with one lag in
.02
this study has 64 parameters and 79 observations. However, estimating separate system of equations from one to 11 lags revealed that the
core results are qualitatively the same. To minimize the problem of over
parameterization a decision was made to limit the number of lags to one.
Empirical results
.2
.00
.0
-.02
-.2
-.04
-.06
10
-.4
Table 2 reports the coefficient estimates and diagnostic tests for the VAR
variable foreclosure starts have no statistically significant values for the
.00
lag of federal funds rate variable using the conventional significance lev-
-.01
els. Moreover, in the higher lag model the signs of the coefficients at
-.02
first, third, fifth, seventh, ninth, and eleventh lags are negative but not
-.03
significant suggesting that an increase in the federal funds rate does not
10
.10
.05
.00
-.05
10
-.10
.04
.10
.02
.05
.00
.00
-.02
-.05
-.04
to capture the true dynamics of the data generating process, and are
-.10
10
.15
.02
.01
.03
10
-.06
10
therefore better suited for statistical inference.7 Thus, the findings from
the impulse response functions are utilized to discuss empirical findings.
Figure 3 plots the impulse response results from the VAR model. The
upper left graph plots the response of the home mortgage interest rates
playing persistence. The upper right corner plots the response of the federal funds rate to the foreclosure starts (FORCST). The finding of a lagged
statistically significant negative response is consistent with the behavior
of the Fed and therefore consistent with the view that as foreclosures
Explained variable
FEDFUND
LOG(HOMIRATE)
FEDFUND(-1)
1.045
(0.047)
0.082
(0.024)
0.003
(0.002)
0.012
(0.01)
HOMIRATE(-1)
-0.292
(0.108)
0.794
(0.056)
-0.008
(0.005)
-0.005
(0.022)
LOG(PRICE(-1))
-0.218
(0.336)
-0.199
(0.175)
0.991
(0.015)
0.128
(0.067)
LOG(FORCST(-1))
-0.650
(0.195)
-0.119
(0.101)
-0.008
(0.009)
0.960
(0.039)
2.421
(2.332)
2.024
(1.211)
0.096
(0.103)
-0.715
(0.465)
0.966
0.969
0.996
0.956
with the view that a once and for all rise in the federal funds rate would
600.785
672.298
5326.526
456.608
Akaike AIC
1.310
0.000
-4.923
-1.914
Schwarz SC
1.449
0.138
-4.784
-1.775
R-squared
F-statistic
LOG(PRICE)
LOG(FORCST)
increase, the Federal Reserve responds by cutting the federal funds rate.
Explanatory
variable
The middle left graph plots the response of home purchase prices (PRICE)
to the federal funds rate. The lagged negative but statistically insignificant
finding may suggest that even though there is a tendency for prices to fall
in response to a Fed rate increase, the changes are not strong enough to
generate a statistically significant response. The middle right graph is the
core finding which plots the response of the foreclosure starts to the federal funds rate. Unlike the expectation of a positive response which would
federal funds rate, the response is negative but insignificant, consistent
6 The results from estimating separate system of equations from two to 11 lags are available
from the author upon request.
7 The model was estimated by cutting the sample short at third quarter of 2007 to exclude
the recent financial and economic crisis and the results were qualitatively identical.
59
Explained variable
.20
.15
.10
Explanatory
variable
D(FEDFUND)
D(HOMIRATE)
D(LOG(PRICE))
D(LOG(FORCST))
.08
D(FEDFUND(-1))
0.623
(0.092)
0.100
(0.056)
0.008
(0.005)
0.002
(0.023)
D(HOMIRATE(-1))
0.081
(0.180)
0.225
(0.109)
-0.024
(0.009)
-0.008
(0.045)
D(LOG(PRICE(-1)))
3.179
(1.919)
3.907
(1.158)
0.174
(0.099)
-0.426
(0.480)
D(LOG(FORCST(-1)))
0.070
(0.425)
0.169
(0.257)
0.044
(0.022)
-0.141
(0.106)
-0.052
(0.045)
-0.077
(0.027)
0.008
(0.002)
0.023
(0.011)
0.024
.04
.05
.00
.00
-.05
.12
-.04
1
10
10
.20
.20
.15
.15
.10
.10
.05
.05
.00
.00
R-squared
0.456
0.236
0.128
-.05
-.05
F-statistic
19.032
7.018
3.352
0.564
-.10
-.10
Akaike AIC
0.917
-0.092
-5.009
-1.854
Schwarz SC
1.050
0.041
-4.875
-1.721
10
10
for each region to utilize the highest degrees of freedom instead of just
one VAR system that includes the four regions in one model. Datastream
reports data from four regions: West, South, Northeast, and North Central. The results are indeed qualitatively similar to findings from estima-
The lower left graph plots the response of the foreclosure starts to home
naturally conjecture that because the federal funds rate is not able to
rate for the West, South, Northeast, and North Central respectively. Con-
influence the foreclosure starts, home mortgage interest rates which, ac-
sistent with the findings from the national variables, none of the figures
cording to theory, follow the federal funds rate movements do not have
standard deviation increase in the federal funds rate. Thus the regional
findings are consistent with the national findings providing further sup-
Lastly, the lower right figure plots the response of the home mortgage
port for the view that the federal funds rate appears to be disconnected
interest rates to foreclosure starts. Consistent with the findings from the
upper right corner that plots the response of the federal funds rate to the
foreclosure starts, the response of home mortgage interest rates to the
The series in the form of levels contain long memory while in the form of
system, an old shock to the series has virtually no effect on the current
mostly fixed rate borrowing rather than adjustable rate but the effect is not
value of the series, if the shock happened long enough ago. For a long
channeled back in the form of lower foreclosures rates. This is perhaps due
memory series, an old shock will still have a noticeable impact on the
to the postulated muted effect coming from the federal funds rate. Fur-
current value of the series. The system estimated in levels captures long
ther, when the upper left graph is compared with the lower right graph, the
memory. To capture the short memory dynamics, the series are trans-
home mortgage interest rate appears to respond with a larger lag to move-
ments in the foreclosure starts than movements to the federal funds rate.
To conduct more in-depth analysis of the data, we use region specific
housing indicators to check consistency with the estimation results from
60
8 Some practitioners in the field intuitively argue that the home mortgage interest rates are
more responsive to 30 year bond yields.
9 To save space only the impulse responses from the core findings are reported in Figure 4.
The remaining impulse responses are available from the authors.
.15
.03
.08
.2
.10
.02
.04
.1
.05
.01
.00
.00
.0
.00
-.01
-.05
10
-.1
10
.03
10
.02
.08
.01
.04
.00
.00
-.04
2
10
10
.15
.10
.05
.05
.00
-.05
-.05
-.10
10
10
-.10
10
.12
.08
.03
.02
.01
In a short memory series, an old shock to the series has virtually no effect
.00
on the current value and thus the shock must have happened a short
.04
.00
-.04
.15
.00
-.04
.20
.10
-.01
.20
.3
10
-.01
10
while ago for the series to respond immediately. Indeed, one would observe in real life that the Federal Reserve response to foreclosures is not
of immediate but of more lagged nature, say after a series of incoming
short run dynamics. Consistent with the findings from the VAR system in
starts to the federal funds rate for the West, South, Northeast, and North
levels, the lag of the differenced federal funds rate does not have a statis-
mic first differences. Consistent with the previous findings, none of the
lag of the foreclosures together with the price series do appear to have a
to a one standard deviation increase in the federal funds rate. The regional findings from a system in first differenced variables again produce
levels. This provides further evidence in favor of the view that changes in
home mortgage interest rate to federal funds rate is significant but much
the federal funds rate do not appear to have a statistically significant im-
short lived with three quarters lag. As expected, the persistence that ap-
pact on foreclosure starts. This finding is consistent with the view that the
pears in the VAR system in log levels which captures long memory seems
ences that capture short memory. The core finding here, the middle right
graph, is consistent with the results from the prior specification. In other
words, when the system is estimated in differenced form, there appears
to be no statistically significant response of foreclosures to the federal
funds rate either. The difference in these results is the response of the
federal funds rate to foreclosures not being statistically significant. So is
the response of the home mortgage interest rates to foreclosures. These
two results were found to be significant in the form of logarithmic levels.
10 The own response of the foreclosures to a standard deviation increase in foreclosures was
significant in all the VAR estimations providing strong support for this claim. Because other
impulse responses provided similar results, these figures are not reported in this study to
save space, but they are available from the authors.
11 To test whether the results are model specific, an ARCH test was conducted on all
variables. None of them appeared to exhibit an ARCH process after the LM test was not
statistically significant at the 5 and 10% significance levels.
61
Coefficient
Std. Error
t-Statistic
Prob.
0.118
0.099
1.191
0.237
DU2
0.624
0.796
0.784
0.435
SLD1
-0.002
0.005
-0.415
0.679
SLD2
-0.053
0.019
-2.777
0.007
SLDP1
0.001
0.012
0.059
0.953
SLDP2
0.011
0.044
0.257
0.798
SLDU1
0.000
0.000
1.642
0.104
SLDU2
0.000
0.002
-0.100
0.920
able. We test for structural breaks and include a dummy in the determin-
FORCST(-1)
0.534
0.112
4.740
0.000
R-squared
0.970
Adjusted R-squared
0.967
Durbin-Watson stat
1.835
DU1
Concluding remarks
Several academicians and practitioners stated in the past that the channel effect in the US is relatively muted. These statements have been
merely conjecture with no extensive empirical documentation. This study
possible structural break starting from the third quarter of 2006. Intercept
tween the federal funds rate and housing market activity as proxied by
dummy variables are represented by (DU1 and DU2) whereas the slope
dummies are represented by (SLD1, SLD2) for the federal funds rate,
VAR models utilizing aggregate and regional data to find evidence con-
(SLDP1 and SLDP2) for the home mortgage interest rate, (SLDU1 and
sistent with the view that the channel effect in the US is relatively muted
SLDU2) for home prices. The first order autoregressive term is repre-
whereby any increase in the federal funds rate does not lead to an in-
before the break but a stronger negative and statistically significant relationship after the break despite the significantly low degrees of freedom.
The results are qualitatively the same with the inclusion of the lagged
dependent variable. They are also the same without the lagged depen-
support for the independent process postulation. These results are also
roles at the micro level are also likely to be more precise than traditional
consistent with the correlations that are run before and after the break.
ones.
rate and reserve requirement ratio have become less effective in terms
References
Barberis, N., A. Shleifer, and R. Vishny, 1998, A model of investor sentiment, Journal of
Financial Economics, 49(3): 307-343
Bollerslev, T., 1986, Generalized autoregressive conditional heteroskedasticity, Journal of
Econometrics, 31: 307-327
Bollerslev, T. and J. Wooldridge, 1992, Quasi-maximum likelihood estimation and inference in
dynamic models with time-varying covariances, Econometric Review, 11: 143-172
Charemza, W. W. and D. F. Deadman, 1997, New Directions in Econometric Practice, London:
Edward Elgar Publishing Limited
Daniel, K., D. Hirshleifer, and A. Subrahmanyam, 1998, Investor psychology and security
market under- and over-reactions, Journal of Finance, 53(6): 2143-2184
DeLong, J. B., A. Schleifer, L. H. Summers, and R.J. Waldmann, 1990, The economic
consequences of noise traders, Journal of Political Economy, 98(4): 703-738
Dickey, D. A. and W. A. Fuller, 1979, Distribution of the estimators for autoregressive time
series with a unit root, Journal of American Statistical Association, 74: 427-431
Dickey, D. A. and W. A. Fuller, 1981, Likelihood ratio statistics for autoregressive time series
with unit root, Econometrica, 49: 1057-1072
Diebold, F.X., 1998, The past, present and future of macroeconomic forecasting, The Journal
of Economic Perspectives, 12: 175-192
Doan, T.,1990, Regression Analysis of Time series, Evanston, IL: VAR Econometrics
Doan, T. and R. Litterman,1986, Users manual RATS:Version 2.0., Evanston IL: VAR
Econometrics
Enders, W.,1995, Applied Econometrics Time Series, New York: John Wiley & Sons
Engle, R. F. and C. W. J. Granger, 1991, Long-Run Economic Relationships: Readings in
Cointegration, New York: Oxford University Press
Fraser, P., M. Hoesli, and L. McAlevey, 2008, House price and bubbles in New Zealand,
Journal of Real Estate Finance and Economics, 37: 71-91
Froyen, R. T., 1996, Macroeconomics, Theories & Policies, fifth edition, New Jersey: PrenticeHall
Hakkio, C. S. and C. S. Morris, 1984, Vector autoregressions: a users guide, Research Working
Paper, Federal Reserve Bank of Kansas City, 84-10
Hardouvelis, G., 1987, Reserve announcements and interest rates: Does monetary policy
matter, Journal of Finance, 42: 407-422
Hamilton, J. D.,1994, Time series analysis, Princeton, NJ: Princeton University Press
Hong, H. and J. C. Stein, 1999, A unified theory of underreaction, momentum trading, and
overreaction in asset markets, Journal of Finance, 54(6): 1839-1885
Johansen, S., 1988, Statistical analysis of cointegration vectors, Journal of Economic
Dynamics and Control, 12: 231-54.
Keynes, J. M., 1936, The General Theory of Employment, Interest and Money, London:
Macmillan
Kyle, A. 1985, Continuous auctions and insider trading, Econometrica, 53(6): 1315-1335
Kwon, C. S. and T.S. Shin, 1999, Cointegration and causality between macroeconomic
variables and stock market returns, Global Finance Journal, 10(1): 71 - 81
Litterman, R.B., 1984, Forecasting with Bayesian vector autoregressions five years of
experience, Staff Report, Federal Reserve Bank of Minneapolis, No.95.
Litterman, R. B. and T. M. Supel, 1983, Using vector autoregressions to measure the
uncertainty in Minnesotas revenue forecasts, Quarterly Review, Federal Reserve Bank of
Minneapolis, 7 (2)
Lui, W. N. Strong, and X. Xu, 1999, The profitability of momentum investing, Journal of
Business Finance & Accounting, 26(9): 1043-1091
Lupoletti, W. M. and R. H. Webb, 1986, Defining and improving the accuracy of macroeconomic
forecasts: contributions from a VAR model, The Journal of Business, 59 (2): 263-285
Rubio, M., 2008, Fixed and variable-rate mortgages, business cycles and monetary policy,
Working Paper, Bank of Spain.
Runkle, D. E., 1987, Vector auto regressions and reality, Journal of Business and Economics
Statistics, 5: 437-442
Shiller, R., 1984, Stock prices and social dynamics, Brookings Papers on Economic Activity,
2: 457-498
Simkins, S., 1995, Forecasting with vector autoregressive (VAR) models subject to business
cycle restrictions, Journal of Forecasting, 11: 569-583
Sims, C. A.,1980, Macroeconomics and reality, Econometrica, 48: 1-49
Sims, C. A., 1996, Macroeconomics and methodology, The Journal of Economic
Perspectives, 10: 105-120
Stevenson, A., V. Muscatelli, and M. Gregory, 1988, Macroeconomic Theory and Stabilisation
Policy, London: Philip Allan
Webb, R. H., 1999, Two approaches to macroeconomic forecasting, Economic Quarterly,
Federal Reserve Bank of Richmond, 85(3), Summer
63
Cutting Edge
66
Cutting Edge
Sovereign Credit
Risk in a Hidden
Markov RegimeSwitching
Framework. Part 2
Louise Potgieter Full time MBA student, Cass Business School, City University
Gianluca Fusai Full Professor in Mathematical Finance, DiSEI University of Piemonte
Orientale
Abstract
markets.
67
Introduction
net fiscal assets, and other assets minus entities too important to fail.
The recent credit crisis has raised the awareness of investors and regula-
tors concerning the appropriate methods for valuation, trading, and risk
ereign default arises when sovereign assets cannot cover the promised
the strike of the above mentioned call option, is therefore defined as the
present value of these payments. While the liabilities are known with a
fair degree of certainty over any given time horizon, the sovereign assets
are more uncertain as assets may change for a large number of reasons.
discussed in Gray et al. (2007) and we allow for structural changes and
Three factors therefore drive default: the sovereign asset value, the vola-
tility of the assets, and the default barrier. The default barrier is defined as
senior foreign-currency denominated debt [Crouhy et al. (2000)].
The model allows for the computation of the term structure of default
probabilities (PDs) and it is calibrated to observed quotes of sovereign
(A) and local-currency liability in foreign currency term (LCL). We use the
international market.
to incorporate the impact of suddenly changing macroeconomic conditions on the sovereign balance sheet information. This is done by assum-
The CCA approach relies on the relationship between balance sheet en-
ing that the volatility of a sovereigns asset value has switching dynamics
chain could represent the states of the economy. The final aim of the re-
cy terms (LCL) is a call option of the sovereigns assets (A) with the strike
balance sheet and to understand if the credit market conveys useful in-
ties according to a CCA model, we perform a reverse engineering procedure by filtering observed market data through the regime-switching
Potgieter and Fusai (2013) provides the key ideas of the model, describ-
model and we try to infer balance sheet information. This allows us to ap-
den Markov Model, and derives the value for a sovereigns assets and a
68
ability (LCL) can be compared with the valuation implied in market quota-
how much CDSs, conditional to the use of the proposed model, can tell
inferring the model parameters, iteratively adjusted to best fit the market
mization. Then we can use the calibrated model to infer the value of the
Gray et al. (2007) extends the Mertons Contingent Claim Approach to the
South Africa, Brazil, US, Italy, and Germany. The analysis covers a window
This section reports the results of the empirical analysis performed in both
The data
calculated by reference to a daily federal funds rate. For Brazil and South
Credit Default Swaps (CDSs) are generally issued for full range of sov-
Africa, interest rates have been obtained using the zero curve as com-
ereign bond issues, more typically for maturities ranging from one to 10
puted by DataStream.
The data source for all CDS quotes and balance sheet data includes both
CDS historical market quote series vary from country to country, depend-
obtainable on a daily basis, given that the sovereign balance sheet data is
The data for South Africa includes CDS market quotes averaged on a
CDS market quote on the previous quarter to have the same frequency
2012. The estimated parameters calibrated across the entire term struc-
ture on the last day of each quarter of the estimation period are presented
in Table 1. The first column is the date when the calibration is performed.
For the purpose of this approach, sovereign default arises when sover-
The second and third column present the volatility in state 1 and 2 re-
spectively, and provide an indication of how the states are defined e.g.,
good or bad. Columns four and five give indication of the probability
the present value of these payments. The default barrier may be defined
the leverage parameter i.e., the ratio of the sovereign asset value and the
denominated debt [Crouhy et al. (2000)]. This research adopts the latter
observed distress barrier. An obligor defaults when the value of its assets
falls below the value of its liabilities, or equivalent when its inverse lever-
policymakers during stress periods. These efforts make such debt more
From the estimated transition probabilities, the probability that the coun-
strike price as the default barrier since such liabilities demonstrate eq-
to state 2 is low. For example, consider the results on June 29, 2012, aj,i
s1
s2
a11
a12
S/K
31/03/2010
85.01%
2.21%
0.9869
0.9891
1.89
30/06/2010
62.97%
3.52%
0.9790
0.9837
1.75
data such as the base money, domestic interest rates, foreign interest
30/09/2010
71.01%
4.15%
0.9853
0.9871
1.92
31/12/2010
70.03%
10.97%
0.9863
0.9874
2.12
31/03/2011
76.19%
3.14%
0.9885
0.9883
1.91
30/06/2011
93.53%
1.46%
0.9704
0.9864
1.98
30/09/2011
86.37%
1.20%
0.9847
0.9842
2.04
Interest rate data include the Eonia (Euro Overnight Index Average) and
30/12/2011
79.21%
2.55%
0.9774
0.9823
2.09
30/03/2012
84.58%
2.48%
0.9839
0.9863
1.91
29/06/2012
99.90%
0.27%
0.9990
0.9865
1.89
lending transactions in the interbank market. It is one of the two benchmarks (the other one being Euribor) used in the money and capital mar-
Table 1: Estimated parameters in two-state hidden Markov regimeswitching model: South Africa
kets in the Eurozone. The US OIS is an interest rate swap that allows
financial institutions to swap the interest rates they are paying without
having to change the terms of contracts in place with other financial institutions. The fixed rate of OIS is considered less risky than the corresponding interbank rate (LIBOR), as only the net difference in interest
rates is paid at maturity of the swap. In the United States, OIS rates are
To state 1
To state 2
From state 1
0.9990
0.0010
From state 2
0.9865
0.0135
69
Calibration
T=1
T=2
T=3
T=4
T=5
T=7
T = 10
This shows that the Markov model captures the persistence in a state;
31/03/2010
1.90%
2.37%
2.19%
2.89%
3.68%
4.53%
5.43%
in this case state 1. How is the state defined then? The two volatility es-
30/06/2010
2.59%
3.07%
2.24%
2.98%
3.84%
4.79%
5.82%
timates provide a clear indication that the level of volatility defines each
30/09/2010
2.10%
2.51%
2.06%
2.73%
3.50%
4.34%
5.25%
state. Given that s1 >> s2, state 1 and state 2 therefore could be classified
31/12/2010
2.06%
2.39%
1.88%
2.53%
3.36%
4.37%
5.54%
31/03/2011
1.40%
1.84%
1.95%
2.87%
3.83%
4.83%
5.85%
30/06/2011
1.48%
2.09%
1.49%
2.68%
3.90%
5.15%
6.43%
30/09/2011
2.70%
3.29%
3.07%
3.97%
4.98%
6.09%
7.27%
30/12/2011
2.73%
3.27%
2.55%
3.52%
4.61%
5.80%
7.06%
30/03/2012
1.82%
2.38%
2.35%
3.37%
4.46%
5.60%
6.77%
29/06/2012
1.75%
2.45%
3.50%
4.64%
5.83%
7.05%
8.30%
Table 3 Calibrated term structure of default probabilities using regimeswitching model: South Africa
70
to global financial stability eased off as the economic recovery gained steam.
Distress
Barrier
Implied
Asset
Value
LCL
Observed
LCL
Estimate
(T=5)
Average
LCL
Estimate
31/03/2010
316,610
599,827
64,526
66,600
65,144
Heavy public debt burdens and weak growth prospects in many advanced
30/06/2010
334,956
584,993
62,147
63,232
62,314
30/09/2010
311,111
597,524
75,756
77,231
76,144
were the culprits. Despite higher growth prospects for emerging economies,
31/12/2010
299,561
634,429
81,653
82,607
81,807
markets faced the risk of sharp reversals. At this stage, the crisis moved into
31/03/2011
311,863
596,100
80,162
84,476
82,443
its fifth year, entering into a new phase in which political differences across
30/06/2011
318,272
631,165
83,871
87,120
85,209
30/09/2011
383,228
782,489
76,628
81,232
79,280
30/12/2011
390,675
814,713
82,327
85,226
83,672
30/03/2012
375,333
718,745
82,077
88,584
86,039
shift to safety and liquidity. Net capital flows to emerging markets such as
29/06/2012
409,865
773,502
81,336
95,011
90,597
South Africa remained relatively strong during the first half of 2011 although
Table 4 Implied sovereign asset value, distress barrier and localcurrency liability (in millions USD): South Africa
In September 2011, the Global Financial Stability Report cautioned that the
risks to global financial stability increased substantially in prior months.
very volatile. This reflected higher nominal interest rates, the perception that
currencies would appreciate, and relatively strong fundamentals.
900.000
800.000
tory power for the observed local-currency liability (LCL) over a T= 5 year
700.000
maturity, using the estimated LCL obtained from the model. For this pur-
600.000
500.000
400.000
300.000
Distress barrier
is as follows:
200.000
100.000
05/2012
03/2012
11/2011
01/2012
09/2011
07/2011
05/2011
03/2011
01/2011
11/2010
09/2010
07/2010
05/2010
LCLObserved = + b LCLEstimate +
03/2010
(1)
Here represents the intercept of the linear regression, b the slope and
, the error term reflecting other factors that influence the observed local-
90.000
85.000
Coefficients
Standard Error
t Stat
P-value
17644.63
8123.17
2.1721
0.061613
0.7322
0.09948
7.3605
7.91E-05
80.000
75.000
70.000
65.000
LCL observed
60.000
R = 0.87
05/2012
03/2012
01/2012
11/2011
09/2011
07/2011
05/2011
03/2011
01/2011
11/2010
09/2010
07/2010
05/2010
50.000
03/2010
55.000
the sovereign balance sheet data required for the contingent claim ap-
gression. This suggests that the LCLEstimate implied in the CDS quotes
has high power in explaining the LCLObserved. On the other hand, given
the estimated value of b lower than 1 due to the fact that LCLEstimate
> LCLObserved, the prediction appears biased i.e., CDS market quotes
The LCL component is reported for both a T = 5 year maturity and the
Table 6 extends the analysis using the LCL estimates implied by different
points of the term structure of default probabilities. The short term maturities can provide an even greater predictive value of the observed value of
LCL. In addition, the prediction extracted from the one year maturity also
appears unbiased ( = 0,b =1). The result implies that the estimated LCL
well above the distress value for all quarters indicating that no default
tend to capture the level of the observed balance sheet data very well.
Maturity
T=1
T=2
T=3
T=4
T=5
T=7
T = 10
This is also confirmed by the inverse leverage ratio that is well above 1 as
R2
0.99
0.99
0.92
0.90
0.87
0.84
0.80
we have seen from Table 1. The value for the LCL estimate for five-year
0.00
275.05
11059.22
14451.81
17644.63
22212.72
29537.93
0.99
0.98
0.86
0.79
0.73
0.68
0.61
a reasonable range, signalling the reliability of the proposed reverse engineering procedure.
71
A case of Brazil
In the case of Brazil, the estimation period ranges from June 2005 to
June 2012. The estimated parameters calibrated across the entire term
structure are presented in Table 7.
The high values of a11 indicate a clear persistence in state 1, which according to the volatility estimates, is classified as a low volatility state.
The inverse leverage ratio varies between 1.5 and 2.4 throughout the
estimation period, suggesting that in this period, Brazil was in quite a
comfortable situation.
Figure 4 shows the term structure of estimated probability of defaults
(PDs) obtained via the regime-switching model. Default probabilities have
remained relatively stable for shorter maturities while the perceived risk
of longer horizons has steadily decreased over the estimation period:
Figure 4 Calibrated term structure of default probability: Brazil
72
s1
s2
a11
a12
S/K
the probability of default for the 10-year maturity steadily declined from
30/06/2005
24.21%
18.65%
0.9990
0.9990
1.8290
above 15% to below 10%, fluctuating at a moderate level with the eco-
30/09/2005
0.81%
7.95%
0.9990
0.9674
1.5049
30/12/2005
22.65%
19.52%
0.9990
0.9990
2.0049
31/03/2006
0.81%
8.76%
0.9986
0.9859
1.5035
30/06/2006
0.80%
11.30%
0.9970
0.9990
1.4981
Table 8 presents the sovereign balance sheet data required for the con-
29/09/2006
0.80%
9.34%
0.9990
0.9899
1.5017
tingent claim approach (CCA) and the local-currency liabilities LCL com-
29/12/2006
0.80%
9.60%
0.9990
0.9990
1.4982
ponent, observed and estimated for a T= 5 year maturity and the coun-
30/03/2007
0.81%
8.38%
0.9986
0.9926
1.5013
29/06/2007
0.81%
9.18%
0.9990
0.9990
1.4985
28/09/2007
0.81%
7.27%
0.9990
0.9883
1.5027
31/12/2007
0.80%
8.98%
0.9990
0.9919
1.5003
31/03/2008
0.80%
11.50%
0.9990
0.9981
1.5015
nomic conditions in Brazil are evident by the rise in the indicators. The
30/06/2008
0.80%
7.18%
0.9990
0.9831
1.5022
30/09/2008
0.68%
6.59%
0.9854
0.9668
1.4369
31/12/2008
0.75%
21.05%
0.9756
0.9700
1.2733
thereafter. This implies that CDS quotations, differently from South Africa,
31/03/2009
0.80%
22.04%
0.9829
0.9829
1.2643
at first overstate and then understate local sovereign liabilities. The dis-
30/06/2009
0.80%
11.49%
0.9988
0.9989
1.4965
tress barrier remains well below the implied asset value aside from March
30/09/2009
0.82%
8.03%
0.9990
0.9835
1.5013
31/12/2009
0.82%
7.38%
0.9990
0.9827
1.5025
31/03/2010
0.86%
7.79%
0.9990
0.9828
1.5022
around this time period, when the US stock market reached its lowest
30/06/2010
3.38%
8.90%
0.9990
0.9803
1.6681
30/09/2010
1.25%
10.03%
0.9990
0.9846
1.7390
31/12/2010
3.88%
11.59%
0.9990
0.9879
1.7863
Next, the quarterly predictive ability for the observed local-currency li-
31/03/2011
9.24%
14.71%
0.9990
0.9912
2.0130
30/06/2011
8.12%
19.37%
0.9990
0.9990
2.2920
liability obtained from the model. The regression analysis provides the
30/09/2011
0.81%
19.85%
0.9860
0.9859
1.2963
30/12/2011
0.80%
11.54%
0.9989
0.9985
1.4964
30/03/2012
0.81%
8.16%
0.9990
0.9851
1.5015
29/06/2012
0.85%
8.96%
0.9990
0.9838
1.5018
Distress
Barrier
Implied
Asset
Value
LCL
Observed
LCL
Estimate
(T=5)
Average
LCL
Estimate
1.400.000
30/06/2005
191,309
572,777
275,667
303,683
296,816
1.000.000
30/09/2005
183,151
648,459
300,826
339,794
333,364
30/12/2005
169,450
591,055
344,974
397,900
391,717
166,652
625,321
349,438
413,395
407,774
30/06/2006
156,661
564,884
369,277
443,110
437,896
29/09/2006
159,560
598,639
362,863
440,745
434,799
29/12/2006
172,589
681,523
388,391
473,197
467,351
30/03/2007
182,082
675,851
402,677
496,021
490,100
29/06/2007
191,358
882,475
432,881
531,538
525,733
28/09/2007
195,331
749,241
501,279
594,251
588,641
31/12/2007
193,219
757,899
591,676
601,720
596,871
31/03/2008
201,637
740,365
613,378
569,518
564,815
30/06/2008
205,536
754,038
696,894
573,100
569,530
30/09/2008
211,381
732,434
725,278
549,452
544,967
800.000
600.000
400.000
Distress barrier
200.000
0
06/2005
12/2005
06/2006
12/2006
06/2007
12/2007
06/2008
12/2008
06/2009
12/2009
06/2010
12/2010
06/2011
12/2011
06/2012
31/03/2006
1.200.000
1.250.000
1.050.000
850.000
198,340
597,776
518,923
413,486
410,399
250.000
31/03/2009
192,676
420,082
484,679
242,094
238,810
30/06/2009
198,996
712,512
553,688
524,346
521,991
30/09/2009
204,934
786,536
699,871
591,623
589,518
31/12/2009
198,192
769,321
765,835
579,311
577,473
31/03/2010
211,532
794,666
728,538
592,566
590,502
30/06/2010
228,649
888,302
781,736
670,571
668,290
30/09/2010
247,812
933,830
813,936
696,602
694,517
31/12/2010
256,804
924,062
876,930
677,673
675,660
31/03/2011
275,947
1,086,314
922,448
822,211
819,955
LCLEstimate
30/06/2011
291,648
1,143,223
1,018,431
862,636
860,418
2
R = 0.81
30/09/2011
298,219
1,055,958
966,011
773,238
769,524
30/12/2011
298,204
1,193,837
874,191
833,793
830,632
30/03/2012
289,606
1,060,937
900,813
784,285
781,434
29/06/2012
287,529
1,017,470
832,855
745,029
741,555
Table 8 Implied sovereign asset value, distress barrier and localcurrency liability (in millions USD): Brazil
LCL observed
450.000
50.000
06/2005
11/2005
04/2006
09/2006
02/2007
07/2007
12/2007
05/2008
10/2008
03/2009
08/2009
01/2010
06/2010
11/2010
04/2011
09/2011
02/2012
31/12/2008
650.000
Coefficients
Standard Error
t Stat
P-value
-89857.77
68743.39
-1.3072
0.202186
1.2518
0.1161
10.7827
2.75E-05
T=1
T=2
T=3
T=4
0.89
0.87
0.85
0.91
26229.31
-4255.86
-34375.15
-63074.92
1.07
1.11
1.16
1.21
T=5
T=7
T = 10
0.81
0.87
0.82
1.34
1.48
The explanatory power of the estimated local-currency liability as measured by the R2 across the entire term structure is presented in Table 10.
R2
Its high value across maturities implies that the estimated LCL tends to
capture the variation in the observed balance sheet data. For Brazil as
low bias.
73
A case of Italy
Table 11 provides the calibrated parameters for Italy over a time period
that extends from June 2008 to June 2012.
The high transition probability a11 of remaining in state 1 indicates a clear
persistence in state 1 which according to the volatility estimate is clearly
a high volatility state. The time pattern of the ratio S/K signals that Italy
went through a very critical period up to June 2011 followed by the start
of a recovery. This can also be appreciated by observing the calibrated
term structure of the default probabilities in Figure 7.
What is observable is how perceived risk at a shorter time horizon jumped
after April 2011, easing off towards the end of the estimation period. During this time Berlusconis government was in conflict over budget cuts
and the default probability of the one-year horizon reached almost 8%
from lows of 2%. Generally, the perceived riskiness of Italian sovereign
The most recent risk profile shows a hump: default probabilities first de-
no default occurring until 2017). This means that markets are expecting
dampened reduced risk of default five years from June 2012 probably
agenda, and for steering the country off the cliff of default. This is evident
in our results: the volatility in the bad state declined steeply from 100%
reform. But over the 10 year term they are anticipating greater risk. Inci-
to a 40% low.
dentally, Moodys rating agency downgraded Italys rating by two notches and subsequently downgraded 10 Italian banks in July 2011. Moodys
cited the usual fears related to the Eurozones debt crisis, along with the
increasing likelihood that greater collective support will be needed for
74
s1
s2
a11
a12
S/K
30/06/2008
99.90%
1.42%
99.90%
99.75%
1.108532
30/09/2008
99.90%
0.40%
99.90%
99.70%
1.028349
31/12/2008
99.90%
0.39%
99.90%
99.75%
1.018536
31/03/2009
99.14%
0.11%
99.90%
99.43%
1.006850
30/06/2009
99.61%
0.25%
99.90%
99.57%
1.020102
30/09/2009
99.90%
0.10%
99.90%
99.70%
1.004666
31/12/2009
99.88%
0.12%
99.90%
99.47%
1.007596
31/03/2010
99.90%
0.10%
99.90%
99.66%
1.005046
30/06/2010
97.77%
1.50%
99.90%
99.90%
1.067598
30/09/2010
95.38%
0.37%
99.90%
99.46%
1.019305
31/12/2010
72.75%
0.20%
99.90%
98.98%
1.011849
31/03/2011
99.90%
0.10%
99.90%
99.71%
1.003614
30/06/2011
92.44%
0.33%
99.90%
99.14%
1.024536
30/09/2011
85.37%
5.80%
99.90%
99.26%
1.275628
30/12/2011
99.90%
11.11%
99.06%
99.77%
1.553669
30/03/2012
41.69%
24.12%
67.70%
79.37%
2.166822
29/06/2012
42.33%
35.93%
54.34%
66.90%
3.086906
Distress
Barrier
2.500.000
729,202
808,342
222,849
260,498
239,305
770,057
277,605
249,385
221,367
31/12/2008
708,299
721,431
318,999
226,554
202,541
31/03/2009
753,850
759,014
343,891
256,392
226,451
Distress barrier
30/06/2009
777,003
792,621
307,652
228,342
203,217
30/09/2009
816,334
820,146
326,721
268,907
235,747
31/12/2009
790,676
796,685
287,591
173,806
154,956
31/03/2010
838,388
842,622
361,719
280,812
247,478
30/06/2010
830,750
886,909
366,995
277,538
255,028
30/09/2010
836,599
852,754
356,730
285,615
254,903
31/12/2010
811,807
821,427
312,449
186,951
169,003
31/03/2011
812,795
815,729
355,788
278,443
242,821
30/06/2011
824,494
844,727
328,732
257,137
229,105
30/09/2011
746,986
952,878
388,206
473,777
448,480
30/12/2011
673,391
1,046,227
398,073
478,492
459,768
30/03/2012
677,325
1,467,641
434,547
546,608
515,622
29/06/2012
677,325
2,090,841
387,910
485,534
464,156
06/2012
12/2011
03/2012
09/2011
06/2011
03/2011
12/2010
09/2010
06/2010
03/2010
12/2009
09/2009
06/2009
03/2009
12/2008
09/2008
Average
LCL
Estimate
748,828
500.000
06/2008
LCL
Estimate
(T=5)
30/06/2008
1.500.000
LCL
Observed
30/09/2008
2.000.000
1.000.000
Implied
Asset
Value
600.000
500.000
400.000
Table 12 Implied sovereign asset value, distress barrier and localcurrency liability (in millions USD): Italy
300.000
200.000
LCL observed
100.000
Coefficients
Standard Error
t Stat
233434.82
24030.98
9.7139
7.31E-08
0.3467
0.07373
4.7023
0.000283
06/2008
09/2008
12/2008
03/2009
06/2009
09/2009
12/2009
03/2010
06/2010
09/2010
12/2010
03/2011
06/2011
09/2011
12/2011
03/2012
06/2012
LCLEstimate
P-value
R2 = 0.60
Maturity
T=1
T=2
T=3
T=4
T=5
T=7
T = 10
R2
0.47
0.50
0.53
0.56
0.60
0.64
0.69
0.21
0.25
0.30
0.35
0.46
0.69
Next, the quarterly predictive power for the observed local-currency li-
ability (LCL) over a T = 5 year maturity is examined in Table 13, using the
estimated local-currency liability obtained from the model. The regression analysis produces the following estimates:
sults imply that, differently from the countries considered previously, the
model used to estimate the LCL tend to improve in its ability to capture
The value of R2, the one-tailed t-statistic of 4.7 and the p-value < 0.05,
the variation in the observed balance sheet data as the maturity of the
75
a wave-like behavior of the PD along the date axis, holding maturity constant, particularly pronounced around December 2008 and July 2011.
On 15 June 2011, Dow Jones reported that the one-year CDS spread for
the United States was at 43 basis points (higher than the 41 basis points
spread for Brazil), and that the cost of insuring one-year US debt against
turity and the LCL averaged across the term structure are reported in
default had risen on the back of concerns related to the debt ceiling. The
Table 16.
potential of sovereign default or restructuring and concerns that Eurozone fiscal pressures could spread is reflected in the rise in perceived
Table 16. The value for the local-currency liability (LCL) estimate lies above
the observed value indicating that the regime-switching method tends to
overestimate the observed LCL. This means that sovereign CDS quotations on US debt tend to overstate local sovereign liabilities. The implied
76
s1
s2
a11
a12
S/K
sovereign asset value lies well above the distress value for all quarters sug-
30/06/2008
99.90%
0.10%
99.00%
99.76%
1.5847
gesting that no default was imminent in the estimation period. The peak
30/09/2008
81.41%
0.15%
98.57%
99.60%
1.6992
and sudden drop in the implied asset value in 2011 highlight the sensitivity
31/12/2008
72.04%
8.66%
99.90%
99.62%
1.6530
31/03/2009
71.35%
8.03%
99.90%
99.63%
1.6572
At this point, the leverage declined, volatility of the model fell sharply, and
30/06/2009
99.90%
0.10%
99.18%
99.65%
1.5893
30/09/2009
99.74%
1.91%
99.37%
99.76%
1.5648
31/12/2009
67.95%
8.74%
99.90%
99.80%
1.6565
Next, the quarterly predictive powers for the observed local-currency li-
31/03/2010
69.61%
8.57%
99.90%
99.75%
1.6529
30/06/2010
68.13%
8.46%
99.90%
99.78%
1.6580
30/09/2010
82.98%
9.67%
99.13%
99.64%
1.6361
31/12/2010
56.23%
0.10%
98.23%
99.21%
1.9686
31/03/2011
92.03%
6.46%
99.06%
99.67%
1.5511
30/06/2011
81.00%
8.63%
99.90%
99.88%
1.5832
30/09/2011
34.52%
0.10%
98.83%
98.88%
2.3708
The high value of R2 suggests that the regression model of the estimated
30/12/2011
73.38%
8.02%
99.90%
99.84%
1.6024
30/03/2012
99.90%
0.10%
99.24%
99.59%
1.5931
29/06/2012
71.40%
9.33%
99.78%
99.66%
1.6570
than observed.
Distress
Barrier
Implied
Asset Value
LCL
Observed
LCL
Estimate
(T=5)
Average
LCL
Estimate
Coefficients
Standard Error
t Stat
-654618.21
358446.60
-1.8263
0.8778
1.1417
0.1543
7.3980
2.23E-06
30/06/2008
21,590,625
34,213,671
741,504
1,828,238
1,640,256
LCLEstimate
30/09/2008
19,204,425
32,633,081
834,246
1,782,926
1,622,492
R2 = 0.79
31/12/2008
19,063,457
31,511,456
1,549,762
1,619,281
1,567,835
31/03/2009
17,755,214
29,423,887
1,580,922
1,878,559
1,818,510
30/06/2009
18,880,466
30,006,668
1,626,160
1,842,701
1,793,222
30/09/2009
20,014,068
31,317,414
1,752,930
1,933,729
1,889,990
31/12/2009
19,730,396
32,682,887
1,952,025
2,121,046
2,075,346
31/03/2010
18,802,148
31,077,619
2,033,102
2,287,998
2,237,842
30/06/2010
30/09/2010
17,112,212
19,624,284
28,371,244
32,108,233
1,977,475
1,925,069
2,211,521
2,168,214
2,252,161
2,181,921
2,223,695
P-value
Maturity
T=1
T=2
T=3
T=4
T=5
T=7
T = 10
R2
0.97
0.98
0.97
0.95
0.79
0.64
0.15
473684.52
0.82
1.06
1.14
1.14
0.88
0.18
31/12/2010
19,349,739
38,091,916
1,970,157
2,283,386
31/03/2011
20,987,132
32,552,364
2,269,313
2,492,781
2,430,674
30/06/2011
21,338,182
33,782,801
2,503,646
2,597,764
2,562,947
30/09/2011
20,031,240
47,489,583
2,589,661
2,786,039
2,751,102
ability as measured by the R2 across the entire term structure. The results
30/12/2011
19,506,948
31,258,363
2,598,679
2,820,027
2,789,482
30/03/2012
20,220,053
32,211,718
2,647,675
3,038,320
2,948,079
29/06/2012
19,420,762
32,180,281
2,606,872
3,013,467
2,923,294
Table 16 Implied sovereign asset value, distressed barrier and localcurrency liability (in millions EUR): US
A case of Germany
In the case of Germany, the estimation period extends from June 2008
to June 2012. The calibrated parameters across the entire term structure
are presented in Table 19.
50.000.000
45.000.000
The striking difference with respect to the other countries is the low vola-
40.000.000
tility levels in the two states, reaching an upper level of around 25%. The
35.000.000
30.000.000
25.000.000
20.000.000
Distress barrier
15.000.000
high transition probability a11 of remaining in state 1 indicates persistence in a low risk state economy. The inverse leverage parameter ranges
between 1.6 and 2.
10.000.000
0
06/2008
08/2008
10/2008
12/2008
02/2009
04/2009
06/2009
08/2009
10/2009
12/2009
02/2010
04/2010
06/2010
08/2010
10/2010
12/2010
02/2011
04/2011
06/2011
08/2011
10/2011
12/2011
02/2012
04/2012
06/2012
5.000.000
Figure 13 shows the term structure of probability of defaults (PDs) obtained from the estimated PDs. The term structure from the start of the
estimation period until around January 2009 is reasonably flat, indicating
that perceived risk at the one to 10 year horizon is low and constant. After
this period, the charts show how markets have changed their perception of Germanys long-term prospects over the course of the remaining
3.500.000
period. In two years, the probability of a default over the 10 year horizon
3.000.000
climbs from below 2% to above 6% while the one to three year level re-
2.500.000
mained low. In July 2012, German 10-year debt did, in fact, reach record
lows on the back of fear of a double dip recession in the US. In the same
2.000.000
LCL estimate (T=5)
1.500.000
LCL observed
1.000.000
The quarterly time series of the sovereign balance sheet components ac-
cording to the contingent claim approach for T= 5 year maturity and the
06/2008
08/2008
10/2008
12/2008
02/2009
04/2009
06/2009
08/2009
10/2009
12/2009
02/2010
04/2010
06/2010
08/2010
10/2010
12/2010
02/2011
04/2011
06/2011
08/2011
10/2011
12/2011
02/2012
04/2012
06/2012
500.000
77
s1
s2
a11
a12
S/K
30/06/2008
2.17%
14.65%
0.9987
0.9985
2.1548
30/09/2008
9.44%
1.33%
0.9982
0.9974
1.4759
31/12/2008
0.60%
13.03%
0.9962
0.9906
1.6380
31/03/2009
1.43%
21.35%
0.9990
0.9949
1.6084
30/06/2009
1.23%
19.16%
0.9990
0.9975
1.5531
30/09/2009
0.10%
20.93%
0.9990
0.9967
1.9859
31/12/2009
0.21%
7.54%
0.9987
0.9951
1.7625
31/03/2010
4.56%
13.97%
0.9988
0.9943
2.0623
30/06/2010
4.65%
7.36%
0.9978
0.9920
1.6989
30/09/2010
3.33%
15.13%
0.9979
0.9932
1.9507
31/12/2010
0.82%
8.22%
0.9974
0.9898
1.4235
31/03/2011
1.97%
11.64%
0.9969
0.9918
2.0313
30/06/2011
0.86%
6.61%
0.9951
0.9938
1.1380
30/09/2011
8.03%
17.86%
0.9986
0.9922
2.3000
30/12/2011
2.78%
20.74%
0.9950
0.9917
1.4242
30/03/2012
0.10%
20.81%
0.9972
0.9921
1.5405
29/06/2012
2.80%
25.17%
0.9933
0.9878
1.8044
16.000.000
14.000.000
10.000.000
8.000.000
06/2012
04/2012
12/2011
10/2011
02/2012
08/2011
06/2011
04/2011
02/2011
12/2010
10/2010
08/2010
06/2010
04/2010
02/2010
12/2009
10/2009
08/2009
2.000.000
sovereign liabilities.
06/2009
the term structure. This implies that the CDS quotations overstate local
Distress barrier
04/2009
4.000.000
02/2009
12/2008
6.000.000
10/2008
observed LCL remains quite stable over time. In practice, we observe that
08/2008
served in the probability of default term structure. On the other side, the
12.000.000
06/2008
Next, the quarterly predictive power for the observed local-currency liability (LCL) over a T= 5 year maturity is examined from the regression
results obtained in Table 21, using the estimated local-currency liability
900.000
800.000
700.000
600.000
500.000
400.000
300.000
LCL observed
200.000
06/2012
04/2012
02/2012
12/2011
10/2011
08/2011
06/2011
04/2011
02/2011
12/2010
10/2010
08/2010
06/2010
04/2010
02/2010
12/2009
10/2009
08/2009
06/2009
04/2009
02/2009
12/2008
10/2008
08/2008
R2
06/2008
100.000
unbiased.
Conclusion
Distress
Barrier
Implied
Asset
Value
LCL
Observed
LCL
Estimate
(T=5)
Average
LCL
Estimate
30/06/2008
5,629,811
12,131,255
332,105
766,000
714,543
30/09/2008
5,651,401
8,340,809
296,242
638,475
597,816
31/12/2008
4,869,564
7,976,342
314,663
345,243
337,972
31/03/2009
4,780,107
7,688,443
292,322
390,487
373,137
30/06/2009
4,986,103
7,744,088
304,066
369,280
358,438
30/09/2009
5,172,468
10,272,219
336,526
387,671
378,173
31/12/2009
5,248,690
9,250,694
341,680
386,837
376,706
31/03/2010
5,088,215
10,493,610
337,961
398,169
386,766
30/06/2010
4,915,667
8,351,017
329,289
383,128
372,209
30/09/2010
5,044,933
9,840,971
363,932
438,583
422,975
31/12/2010
5,306,088
7,553,110
408,009
504,486
482,842
31/03/2011
5,218,723
10,600,917
415,524
500,186
480,276
30/06/2011
5,556,759
6,323,703
405,900
486,315
466,626
30/09/2011
5,888,272
13,543,026
367,374
520,247
483,898
30/12/2011
5,550,577
7,905,355
352,365
464,870
437,941
30/03/2012
5,647,322
8,699,482
366,366
510,810
475,596
29/06/2012
5,662,888
10,218,330
349,963
524,980
480,244
inputs into the standard option pricing formula in the Merton Contingent
Table 20 Implied sovereign asset value, distressed barrier and localcurrency liability (in millions USD): Germany
Coefficients
Standard
Error
t Stat
P-value
326508.18
43320.09
7.5371
1.78E-06
0.04479
0.08862
0.5054
0.6206
captures the level in the observed LCL for all cases bar Germany. The
LCLEstimate
regression results appear very good, albeit biased given non-zero inter-
R = 0.02
cepts and betas not equal to 1. However, if we consider very short matur-
Maturity
T=1
T=2
T=3
T=4
T=5
T=7
T = 10
R2
0.99
0.51
0.16
0.06
0.02
0.00
0.05
0.00
115522.37
253870.81
301201.01
323006.38
350404.59
367717.41
1.00
0.63
0.24
0.11
0.05
-0.01
-0.04
First we identify that the LCL requirement according to the model is higher than the observed value for the case of South Africa, the US, and Germany. Brazil and Italy, during certain periods require a lower LCL estimate
than is observed in the market. Any under- or over-estimation of the LCL
estimate relative to the observed value implies that the corresponding
CDS market quotes, jointly with the model used, under- or over-estimate
local sovereign liabilities respectively. We also observe the relation to the
estimated inverse leverage ratio and the implied asset value. Inverse leverage ratios for the case of Italy are very low, showing signs of imminent
default prior to 2011, therefore resulting in a low estimated value for the
LCL for the relevant sample period. A low predicted inverse leverage parameter (close to 1) signifies imminent default and given the nature of the
valuation of the LCL, significantly draws the required LCL lower than its
observed counterpart.
In the August 2012 monthly bulletin of the European Central Bank (ECB),
the linkages between structural, financial, and fiscal imbalances are cited
79
to have led to the sovereign debt crisis and the fragmentation of the
the sample size [Orth (2011)]. This could affect the regulatory approaches
to risk management. Basel II states that where limited data are avail-
able, a bank must adopt a conservative bias to its analysis, adding to its
and correct systematic risk prior to the crisis and was equally challenged
this applies equally in the case of the sovereign scenarios that we have
described.
References
The model provided in this research provides a relative valuation framework for contingent claims on a sovereigns asset, estimating with a
degree of caution, the balance sheet requirements necessary to predict
default. In addition, Gray et al. (2007) suggests that the CCA approach
could have implications on the rapidly growing sovereign wealth funds,
particularly for emerging markets governments that have accumulated
large reserves. Once sovereign risk exposures are calculated in a regime-
Brigo, D. and M. Tarenghi, 2005, Credit Default Swap Calibration and Counterparty Risk
Valuation with a Scenario Credit Default Swaps I: No Counterparty Default Risk, Journal of
Derivatives, 8: 29-40
Clement, D., 2012, Interview with Darrell Duffie, The Region [Interview], 15 June
Crouhy, M., D. Galai, and R. Mark, 2000, A comparative analysis of current credit risk models,
Journal of Banking Finance, 24: 59-117
Duffie, D. and M. Thukral, 2012, Redesigning Credit Derivatives to Better Cover Sovereign
Default Risk, Preliminary Draft, Standford University, available at: http://www.darrellduffie.com/
uploads/working/DuffieThukralMay062012.pdf (accessed 31 August 2012.)
quirements that severely limit the scope of application for some models,
and dependence of some models results on distributional assumptions.
Furthermore, that market imperfections exist implies that non-market
Erlwein, C., R. S. Mamon, and T. K. Sui, 2008, The Pricing of Credit Default Swaps under a
Markov-Modulated Mertons Structural Model, Northern American Actuarial Journal, 12(1):
19-46
European Central Bank (ECB), 2012, Eurosystem Monthly Bulletin, August, available at http://
www.ecb.europa.eu/pub/pdf/mobu/mb201208en.pdf (accessed August 27, 2012)
Gray, D., M. Merton, and Z. Bodie, 2007, Contingent Claims Approach to Measuring and
Managing Sovereign Credit Risk, Journal of Investment Management, 5: 5-28
Leland, H., 2004, Predictions of Default Probabilities in Structural Models of Debt, Journal of
Investment Management, 2: 1-28
Liew, C. C. and T.K. Siu, 2010, A hidden Markov regime-switching model for option valuation,
Insurance: Mathematics and Economics, 47: 374-384
Orth, W., 2011, Default Probability Estimation in Small Samples With an Application to
Sovereign Bonds, Seminar of Economic and Social Statistics, University of Cologne
Potgieter, L. and G. Fusai, 2013, Sovereign Credit Risk in a Hidden Markov Regime-Switching
Framework. Part 1: Methodology, The Capco Institute Journal of Financial Transformation, 37:
99-109
Tarashev, N., 2005, An Empirical Evaluation of Structural Credit Risk Models, Working Paper,
Bank of International Settlements
were estimated well enough by implementing the regime-switching parameters in a contingent claims approach. The regression analysis verifies this. Germany proved to be a case where the explanatory power of
the model is weak and the cause is evident on inspection of the implied
Appendix 1
asset value volatility i.e., volatility implied from the leverage parameter
and not calibrated explicitly. Gray et al. (2007) attribute asset volatility to
high levels of foreign exchange price volatility. This highlights the need to
address the relationship and causality of asset value volatility and various
balance sheet risk. If the total value of the assets falls below the level of
risk indicators.
The following issues are left for future research. The research estimates
put option on the underlying assets with the strike equal to the promised
the Markov model when the number of states is two. It would be better,
underlying asset and strike. The following balance sheet identity ensues:
other parameters. Standard approaches applied to samples with few defaults pose some serious drawbacks which lead to a high probability of
underestimating the true default probability (PD). These include the obvi-
80
size observed in sovereign CDS data, any default event leading to a low
foreign reserves, net fiscal assets, and other assets minus entities too
MLC
base money. Sovereign default arises when sovereign assets cannot suf-
ficiently cover the promised payment on foreign currency debt. The de-
rd
rf
Bd
terms
XF
of that loan equal to the expected loss of default is created. The ac-
tion of the lender consists of pure default-free lending and bearing a risk
of default by the sovereign. Risky debt can be viewed as a contingent
Equation (2) and (4) are typically used to calculate the unknown and
rency liabilities have a similar effect on inflation and price changes as the
The main challenge is deriving an estimate for the market value and vola-
tility of sovereign assets. Because these are not directly observable, the
CCA approach relies on the relationship between balance sheet entries
sovereigns assets (A) with the strike price as the default barrier (Bf). An
estimation of the sovereign unknown and unobserved asset value (A) can
(LCL) is a call option of the sovereigns assets (A) with the strike price as
S
K
A =
The underlying risky asset (S) and the strike (K) in the regime-switching
* Bf
(6)
framework equates to the sovereign asset value (A) and threshold barrier
LCL = AN(d1) B f erfT N(d2)
(2)
(Bf). Similarly, the inverse leverage parameter (S/K) equates to the ratio of
the sovereign asset value (A) to the default threshold (A/Bf).
(3)
The use of the Merton-type model requires many balance sheet inputs
and parameters which are not always clearly observed and can some-
times be inaccurate or difficult to obtain. By reverse engineering the valuation of a sovereigns asset value, the model requires substantially less
(4)
market information and adjusts for any structural breaks in the model in
an attempt to improve the fair value estimates of a sovereigns balance
sheet.
The formula for the local currency liability in foreign currency which is
observed directly from market data is defined as
LCL = M + Bd,FC =
where
(MLCerdT+Bd)erfT
XF
(5)
81
82
Cutting Edge
Viewing Risk
Measures as
Information
Wayne Tarrant Department of Mathematics, Rose-Hulman Institute of Technology
Dominique Gugan Paris School of Economics, MSE - CES, Universit Paris1, PanthonSorbonne
Abstract
Regulation and risk management in banks depend on underlying risk measures. In general this is the only purpose that
is seen for risk measures. In this paper we suggest that the
reporting of risk measures can be used to determine the loss
distribution function for a nancial entity. We demonstrate
that a lack of sufcient information can lead to ambiguous
risk situations. We give examples, showing the need for the
reporting of multiple risk measures in order to determine a
banks loss distribution. We conclude by suggesting a regulatory requirement of multiple risk measures being reported
by banks, including specic recommendations.
83
Introduction
In every instance that we can nd, risk measures are used to compute
simply because there is ten times as much money at stake. And most
would acknowledge that having cash on hand makes them feel safer
legislate the amount of cash reserves a bank must hold. But should this
lowing properties:
In this paper we present the thesis that the reporting of a risk measure
The idea behind this denition is that a positive number implies that one
is at risk for losing capital and should have that positive number of a cash
measures, the Expected Shortfall and the Maximum Loss. We show that
reporting any one of these measures can lead to multiple possible loss
say that the company has enough capital to take on more risk or to return
distributions.
Following the old adage that more information is always better, we con-
The risk measure that is most used is the Value at Risk (VaR). Essentially
sider the situation of reporting any two risk measures. Again, we show
the -Value at Risk is that number L so that we can expect the losses to
that there is ambiguity about the loss distribution, no matter which two
measures we choose. We show the same result for three measures and
our position is $100, we would expect to lose more than $100 only 5% of
There is little wonder that people are confused by what the term risk
Since the paper of Markowitz (1952), many have also favored the useful-
ness of diversication. VaR does not account for this preference, and a
simple example shows this. Consider the case where a bank has made
teen denitions are relevant to the situations of banks and securities rms.
two $1 million loans and one $2 million loan, each with a 0.04 probability
The unifying theme for each of the denitions is that risk requires both un-
of default and all pairwise independent. Then the 95%-VaR for each loan
certainty and exposure. If a company already knows that a loan will default,
there is no uncertainty and thus no risk. And if the bank decides not to loan
and make our portfolio out of the two independent $1 million loans,
measure that will account for all the risk that a bank or securities rm might
encounter. Some have objected to the risk measure being a single number,
but there is some support for this idea. Investing is always a binary deci-
This leads Artzner et al. (1997) to term VaR as incoherent. They dene a
sion: either one invests or one chooses not to invest. Thus the argument is
that, given a single number, one should have enough information to decide
whether to invest or not. There have been some general agreements about
84
Virtually everyone would agree that if the payout for an investment is al-
The most commonly used coherent risk measure is the Expected Short-
ways positive (after accounting for the risk-free interest rate), then there is
fall. The -Expected Shortfall is the amount one would expect to lose on
truly no risk of loss in the investment. Many people would say that there is
average, given that one is in a case where the loss is greater than the
1%
0
1
%
2
20
Here, we discuss the weakness of any single measure of risk for risk
0
management strategy.
In the following we are concerned with tails of probability distribution
10
functions, i.e., if (x) is the pdf of a loss distribution, then we are concerned with the region [c,d] with the property that dc (x)dx = 0.05 and
that (x) = 0 for all x > d. In Figure 1, [c,d] is the region of the graph that we
there are probabilities of a loss in each case. We see that the loss dis-
will consider. This is often called the tail of the loss distribution. In Figure
2 and following, we will only see the region [c,d]. Under our notation, the
1
.05
dc x (x)dx. We
also want to mention that the probability density function of the loss is
just the negative of the probability density function for the returns.
glance. However, such loss distributions may occur when risks are highly
concentrated.
In this and the following sections we will also adopt the notation that the
first graph in each figure represents a loss distribution X1, the second
represents a loss distribution X2, the third a loss distribution X3, and so
losses in the real estate market in some American cities. As one house is
other houses to fall in value and become more likely to enter foreclosure.
Although VaR is used by the Basel Committee for determining the capi-
Since community banks often focus their loans in one particular locality,
talization that a bank needs, it turns out that VaR is insufcient for distin-
It is notable that our other two measures can help us distinguish among
5
,
2
20
ES95%(X2) =
= ES95%(X3), so we are unable to tell the difference be3
tween the second and third loss distributions. However, we have that
85
5
%
12
1
%
4
16
2
%
3
2
%
3
20
15
15
1
%
3
1
%
3
30
30
distribution is such that VaR(X1) = 5. For the second and third loss dis-
1%
-5
tributions we know that VaR(X2) = 0 = VaR(X3). And this time the Expected Shortfall sorts out the different distributions because ES(X1) =
2%
5
10
5
, ES(X2) = , and ES(X3) = .
3
3
2
We hope that the geometry of our gures has pointed out some facts
about the different risk measures. In order for us to obtain two distribu-
1%
0
tions with the same 95%-VaR, we only have to begin our loss distribution
5
tails at the same value, making sure that we have 5% of the probability
after that common point. For two distributions to have the same 95%-Expected Shortfall, we must have two geometric gures that have the same
weighted average. And for two gures to have the same Maximum Loss,
Whether one argues that certain loss distributions are likely or unlikely,
we would require a common point as the maximum value for which both
there is a stubborn fact that remains. The VaR cannot distinguish among
loss distributions have a non-zero probability, i.e., they need the same
ending value.
only require the reporting of VaR, this should be a very troubling fact.
perhaps one of the two other measures is a better choice. We will start
we try using two different risk measures. We will attempt to use measures
in pairs to see if they give us a full description of the risk characteristics
of a loss model.
leaves the same issue as before. There are uniform, rising, and falling
probabilities, all of which have the same 95%-Expected Shortfall of 10.
We will start by using both the 95%-VaR and the 95%-ES. Again we are
In this case, the Maximum Loss once again identies the three situations
left with a problem. In Figure 5 we see that the loss distribution X1 is uni-
because ML(X1) = 16, ML(X2) = 15, and ML(X3) = 30. Here we also know
form, while X2 has rising probabilities as losses grow, and X3 is the most
that VaR(X1) = 4, while VaR(X2) = 0 = VaR(X3), where each of the VaR cal-
standard looking probability tail. It is discouraging that the two most fre-
86
Because the ES has not been able to distinguish among very different
95%-ES of 10. With such different loss characteristics sharing the same
VaR and ES, it is astounding that these are the only two measures men-
Maximum Loss. We will again look at the last 5% of the probability dis-
here the Maximum Loss will distinguish among the loss distributions as
two measures.
As we see in Figure 4, the risk models have the same Maximum Loss
Now we try to use the pair of measures that are the 95%-VaR and the
5. Again, there are very different situations of risk here. The rst loss
1
%
2
2%
0
0
1%
0
1%
0
10
10
1%
0
2%
10
10
1%
0
5
0
Figure 8 Four loss distributions with equal 95%-VaRs, 95%-ESs and MLs
1
%
4
-5
15
1
%
3
15
-15
2
%
3
tagion. Then there may be other neighborhoods where most of the indi15
5
5
Shortfall will distinguish the loss distributions as ES(X1) = , ES(X2) =
,
3
2
10
and ES(X3) = .
3
So now we will consider the nal pair, the circumstance of using the
of examples that share the same VaR, ES, and ML. In order for our ex-
Once again the three loss distributions shown in Figure 7 have contrast-
in order to make the VaRs equal. We require identical ending points for
the MLs to be equal. Since the uniform distribution has symmetry, it will
ties, and one of declining probabilities. Yet we have each of the three loss
distributions with 95%-ESs of 5 and with MLs of 15. The 95%-VaR does
(and the same starting and ending points) must also be symmetric in
VaR(X3) = 0.
10
on the interval [0, 10].
n
ent risk situations, we move to the next logical step. Here we look at the
95%-VaR, the 95%-ES, and the Maximum Loss.
nothing special about the 95% level. Our examples would be just as valid
and a ML of 10. The most compelling question here is whether these loss
and then scaled identically in all of the corresponding diagrams. However, this concept of considering different risk levels might also lead us to
enough risk measures so that we can determine the universal risk prole
87
1%
0
1
%
2
7
2
3
2
ES, and 99%-ES, and a Maximum Loss, where this is possible. We have
demonstrated that the reporting of ve risk measures can theoretically
lead to an innite family of possibilities for the actual loss distribution, but
3
%
2
we have also conceded that such loss distributions are unlikely to occur.
1
%
2
ability in loss distributions when one, two, and three risk measures are
reported. Thus, we suggest that banks be required to submit ve risk
3
%
2
1%
Thus, we feel that banking regulations would be much safer if such regulations require the reporting of ve different risk measures.
ent types of common loss distributions has been written with my student,
VaR, the 99%-VaR, the 95%-ES, 99%-ES, and the ML. The different risk
proles demonstrated in Figure 8 are quite odd-looking, but they all have
along with the level and the specic measures. Finally, another paper
that looks at the predictive ability of historical risk measures is also in
The three distributions all have the necessary geometric properties for
progress.
References
Any gure meeting all these criteria would also be indistinguishable from
the rst ve when viewed purely by looking at the values in our vector of ve measures. Clearly several different proles can share all ve
measures.
do not claim that such loss distributions will occur naturally, with the ob-
that ve risk measures should distinguish among differing loss distributions in almost all natural circumstances.
88
Artzner, P., F. Delbaen, J. Eber, and D. Heath, 1997, Thinking Coherently, Risk 10: 68-71
Basel Committee on Banking Supervision (BCBS), 1988, International Convergence of Capital
Measurement and Capital Standards, BIS, Basel, Switzerland
Basel Committee on Banking Supervision (BCBS), 1993, Amendment to the Basel Capital
Accord to Incorporate Market Risk, BIS, Basel, Switzerland
Basel Committee on Banking Supervision (BCBS), 2005, Basel II: International Convergence of
Capital Measurement and Capital Standards: A Revised Framework, BIS, Basel, Switzerland
Bassi, F., P. Embrechts, and M. Kafetzaki, 1996, A Survival Kit on Quantile Estimation,
Working Paper, Department of Mathematics, ETHZ, Zurich, Switzerland
Caillault, C. and D. Gugan, 2009, Forecasting VaR and Expected Shortfall Using Dynamical
Systems: a Risk Management Strategy, Frontiers in Finance, 6: 26-50.
Embrechts, P., C. Klppelberg, and T. Mikosch, 2007, Modelling Extremal Events for Insurance
and Finance, Berlin: Springer
Fama, E., The Behavior of Stock Markets, 1965, The Journal of Business, 38: 34-54
Gugan, D., and B. Hassani, 2009, A modied Panjer algorithm for operational risk capital
calculations, Journal of Operational Risks, 4: 1-20
Hogg, R. and S. Klugman, 1984, Loss Distributions, New York: Wiley
Mandelbrot, B., 1963, The Variation of Speculative Prices, The Journal of Business, 36:
394-414
Markowitz, H., 1952, Portfolio Selection, The Journal of Finance, VII: 77-91.
Rockafellar, R. T. and S. Uryasev, 2002, Conditional Value-at-Risk for General Loss
Distributions, Journal of Banking and Finance, 26: 1443-1471
Stulz, R. M., 1996, Rethinking Risk Management, Journal of Applied Corporate Finance, Fall
1996: 8-24
Yamai, Y. and T. Yoshiba, 2002, Comparative Analyses of Expected Shortfall and Value-atRisk: Their Validity under Market Stress, Monetary and Economic Studies Bank of Japan, 20:
181-238
89
The first page must provide a short abstract, as well as the full names, titles,
that published papers meet the highest standards of thought leadership and
Note that the Journal style does not include numbered sections. We
recommend formatting section titles in bold 14 pt, 12 pt and 10 pt to
Is your article purely abstract? If so, does it picture a world that could exist
in the future?
Can your propositions be backed by a source of authority?
References at the end of the paper should use the following formats.
Books
Managing the Value of Companies, New York: John Wiley & Sons
Structural readjustment
Innovation & new sources of liquidity
e-Books
Leadership
Financial regulations
Managing the Value of Companies, New York: John Wiley & Sons. Kindle
Financial technology
edition
Contributions to collective works
Manuscript guidelines
Manuscripts should be submitted by e-mail directly to Peter Springett:
Ritter, J. R., 1997, Initial Public Offerings, in Logue, D. and J. Seward (eds),
Warren Gorham & Lamont Handbook of Modern Finance, Cincinnati: SouthWestern College Publishing
[email protected].
Journals and periodicals
All submissions must be in US English.
Government publications
Manuscripts must be in Microsoft Word, using Times New Roman font, size10.
Newspaper articles
reproduces all formulae, graphs and illustrations. When graphs are used,
90
Unpublished material
resolution.
We hope that you will join us on our journey of discovery and help
shape the future of finance.
91
Layout, production and coordination: Cypres Daniel Brandt, Kris Van de Vijver
and Pieter Vereertbrugghen
Graphic design: Buro Proper Bob Goor
Photographs: Bob Goor
2013 The Capital Markets Company, N.V.
Prins Boudewijnlaan 43, B-2650 Antwerp
All rights reserved. All product names, company names and registered trademarks in this document remain the property of their respective owners. The
views expressed in The Journal of Financial Transformation are solely those of
the authors. This journal may not be duplicated in any way without the express
92
written consent of the publisher except in the form of brief excerpts or quotations for review purposes. Making copies of this journal or any portion thereof for
any purpose other than your own is a violation of copyright law.
BARUCH COLLEGE
ZICKLIN SCHOOL OF BUSINESS
The largest accredited collegiate school of business in the United States
One of the nations Best Business Schools
Forbes Magazine
NATIONALLY RANKED:
#26Graduate Accounting program
#24Undergraduate Accounting program for a second year
2012 Professors Survey Public Accounting Report
www.baruch.cuny.edu/zicklin
Amsterdam
Antwerp
Bangalore
Bratislava
Charlotte
Chicago
Dsseldorf
Frankfurt
Geneva
Hong Kong
Johannesburg
London
New York
Orlando
Paris
San Francisco
Singapore
Toronto
Washington, D.C.
Zurich
capco.com