Sie sind auf Seite 1von 105

The Journal of

Risk Finance
Focus on emerging markets
Volume 8 Number 4 2007
ISSN 1526-5943
www.emeraldinsight.com
jrf cover (i).qxd 07/08/2007 10:37 Page 1
Access this journal online _________________________ 323
Editorial advisory board __________________________ 324
EDITORIAL
Thoughts on the scientific method:
part 2 frequentist fecklessness
Michael R. Powers______________________________________________ 325
Calibrating asset correlation for Indian corporate
exposures: implications for regulatory capital
Arindam Bandyopadhyay, Tasneem Chherawala and Asish Saha________ 330
Dividend policy and payout ratio: evidence from the
Kuala Lumpur stock exchange
Abdulrahman Ali Al-Twaijry _____________________________________ 349
Debt policy and performance of SMEs: evidence
from Ghanaian and South African firms
Joshua Abor __________________________________________________ 364
Foreign exchange risk exposure of listed companies
in Ghana
Zubeiru Salifu, Kofi A. Osei and Charles K.D. Adjasi _________________ 380
The Journal of Risk
Finance
Focus on emerging markets
Editor
Michael R. Powers
ISSN 1526-5943
Volume 8
Number 4
2007
CONTENTS
The current and past volumes of this journal are available at:
www.emeraldinsight.com/1526-5943.htm
You can also search more than 150 additional Emerald journals in
Emerald Management Xtra (www.emeraldinsight.com)
See page following contents for full details of what your access includes.
Access this journal electronically
Banks risk management: a comparison study of
UAE national and foreign banks
Hussein A. Hassan Al-Tamimi and Faris Mohammed Al-Mazrooei ______ 394
The North Cyprus banking sector: the effect of a
speculative attack on the Turkish Lira
Nil Gu nsel ____________________________________________________ 410
COMMENTARY
Sums of knowledge
Andrew Green_________________________________________________ 422
CONTENTS
continued
As a subscriber to this journal, you can benefit from instant,
electronic access to this title via Emerald Management Xtra. Your
access includes a variety of features that increase the value of
your journal subscription.
How to access this journal electronically
To benefit from electronic access to this journal, please contact
support@emeraldinsight.com A set of login details will then be
provided to you. Should you wish to access via IP, please
provide these details in your e-mail. Once registration is
completed, your institution will have instant access to all articles
through the journals Table of Contents page at
www.emeraldinsight.com/1526-5943.htm More information
about the journal is also available at www.emeraldinsight.com/
jrf.htm
Our liberal institution-wide licence allows everyone within your
institution to access your journal electronically, making your
subscription more cost-effective. Our web site has been
designed to provide you with a comprehensive, simple system
that needs only minimum administration. Access is available via
IP authentication or username and password.
Emerald online training services
Visit www.emeraldinsight.com/training and take an Emerald
online tour to help you get the most from your subscription.
Key features of Emerald electronic journals
Automatic permission to make up to 25 copies of individual
articles
This facility can be used for training purposes, course notes,
seminars etc. This only applies to articles of which Emerald owns
copyright. For further details visit www.emeraldinsight.com/
copyright
Online publishing and archiving
As well as current volumes of the journal, you can also gain
access to past volumes on the internet via Emerald Management
Xtra. You can browse or search these databases for relevant
articles.
Key readings
This feature provides abstracts of related articles chosen by the
journal editor, selected to provide readers with current awareness
of interesting articles from other publications in the field.
Non-article content
Material in our journals such as product information, industry
trends, company news, conferences, etc. is available online and
can be accessed by users.
Reference linking
Direct links from the journal article references to abstracts of the
most influential articles cited. Where possible, this link is to the
full text of the article.
E-mail an article
Allows users to e-mail links to relevant and interesting articles to
another computer for later use, reference or printing purposes.
Structured abstracts
Emerald structured abstracts provide consistent, clear and
informative summaries of the content of the articles, allowing
faster evaluation of papers.
Additional complimentary services available
Your access includes a variety of features that add to the
functionality and value of your journal subscription:
Xtra resources and collections
When you register your journal subscription online, you will gain
access to Xtra resources for Librarians, Faculty, Authors,
Researchers, Deans and Managers. In addition you can access
Emerald Collections, which include case studies, book reviews,
guru interviews and literature reviews.
E-mail alert services
These services allow you to be kept up to date with the latest
additions to the journal via e-mail, as soon as new material
enters the database. Further information about the services
available can be found at www.emeraldinsight.com/alerts
Emerald Research Connections
An online meeting place for the research community where
researchers present their own work and interests and seek other
researchers for future projects. Register yourself or search our
database of researchers at www.emeraldinsight.com/
connections
Choice of access
Electronic access to this journal is available via a number of
channels. Our web site www.emeraldinsight.com is the
recommended means of electronic access, as it provides fully
searchable and value added access to the complete content of
the journal. However, you can also access and search the article
content of this journal through the following journal delivery
services:
EBSCOHost Electronic Journals Service
ejournals.ebsco.com
Informatics J-Gate
www.j-gate.informindia.co.in
Ingenta
www.ingenta.com
Minerva Electronic Online Services
www.minerva.at
OCLC FirstSearch
www.oclc.org/firstsearch
SilverLinker
www.ovid.com
SwetsWise
www.swetswise.com
Emerald Customer Support
For customer support and technical help contact:
E-mail support@emeraldinsight.com
Web www.emeraldinsight.com/customercharter
Tel +44 (0) 1274 785278
Fax +44 (0) 1274 785201
www.emeraldinsight.com/jrf.htm
JRF
8,4
324
The Journal of Risk Finance
Vol. 8 No. 4, 2007
p. 324
#Emerald Group Publishing Limited
1526-5943
EDITORIAL ADVISORY BOARD
Senior members
David K.A. Mordecai
Risk Economics Ltd
Samantha Kappagoda
Caxton Associates LLC
So hnke M. Bartram
University of Lancaster
Lea V. Carty
Lehman Brothers
Bingzheng Chen
Tsinghua University
Rick A. Cooper
CTN Strategic
Christopher L. Culp
University of Chicago
J. David Cummins
Temple University
John Drzik
Mercer Oliver Wyman
David Felsenthal
Clifford Chance
Kenneth A. Froot
Harvard University
Bennett W. Golub
BlackRock
Anne Gron
NERA Economic Consulting
James B. Heaton
Bartlit Beck LLP
Philippe Jorion
University of California, Irvine
Anil Kashyap
University of Chicago
Andrew Kuritzkes
Mercer Oliver Wyman
Christopher K. Li
Dow Chemical
Robert H. Litzenberger
Goldman Sachs
Patrick J. McCarty
Commodity Futures Trading Commission
David C. Shimko
RiskCapital
Martin Shubik
Yale University
Larry Swertloff
Hudson Bay Portfolios
Dimitrios P. Tsomocos
University of Oxford
Larry Y. Tzeng
National Taiwan University
Stanislav Uryasev
University of Florida
Nicholas Wilson
University of Leeds
EDITORIAL
Thoughts on the scientic
method: part 2 frequentist
fecklessness
Michael R. Powers
Temple University, Philadelphia, Pennsylvania, USA
Abstract
Purpose The purpose of this two-part series is to consider the role of the scientic method (SM) in
human understanding, questioning both its consistency in actual practice and its reasonableness as a
system of philosophy and action.
Design/methodology/approach Part 2 considers problems of inefciency and inertia caused by
the SMs collectivist, frequentist orientation.
Findings It is argued that problems caused by the SMs frequentist framework may be avoided by
a more individualist, Bayesian approach.
Originality/value The two-part series challenges certain aspects of the scientic method as
employed in the practice of modern science.
Keywords Scientic management, Philosophy, Bayesian statistical decision theory
Paper type Viewpoint
This editorial is the second of a two-part critique of the scientic method (SM) as it is
commonly practiced. In the rst part (Powers, 2007), I addressed the problem of
inconsistencies in the way the SM is applied, and argued that these inconsistencies
may lead researchers to ignore important phenomena, especially when the phenomena
are of a subtle and/or transient nature. In the present installment, I consider problems
of inefciency and inertia caused by the SMs collectivist, frequentist orientation, and
show how these problems may be avoided by a more individualist, Bayesian approach.
We begin by recalling from part 1 the following conventional outline of the
scientic method:
(1) The development of a null hypothesis (H0) summarizing the relationship to be
tested.
(2) The identication of an observable phenomenon that is predicted by the
underlying hypothesis, and not otherwise explicable.
(3) The design of an experiment comprising a reasonably large number of
individual observations, each of which reects the presence/absence of the
identied phenomenon.
(4) The replication of the experiment by several independent researchers.
Circular reasoning?
Turning to the SMs philosophical underpinnings, one quickly encounters what
appears to be a problem of circular logic. This is because the SM, as a process of human
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1526-5943.htm
Thoughts on the
scientic
method
325
The Journal of Risk Finance
Vol. 8 No. 4, 2007
pp. 325-329
qEmerald Group Publishing Limited
1526-5943
DOI 10.1108/15265940710818940
understanding, is entirely observational in nature, and at the same time, its intellectual
justication seems to arise exclusively from its observed successes. In other words, to
place credence in the SM, one must already accept the SM as a means of generating
belief.
This apparent shortcoming is a fundamental problem of British empiricism, the
philosophy that the SM formalizes. Developed by the late renaissance writers John
Locke, George Berkeley, and David Hume, empiricism argues that human knowledge
derives from the observation of objects in the external world through the bodys senses
(induction), and that internal reection is useful primarily in identifying logical
relationships among these observables. In his famous treatise, An Enquiry Concerning
Human Understanding (1748), Hume acknowledges the issue of circularity as follows:
We have said that all arguments concerning existence are founded on the relation of cause
and effect; that our knowledge of that relation is derived entirely from experience; and that all
our experimental conclusions proceed upon the supposition that the future will be
conformable to the past. To endeavor, therefore, the proof of this last supposition by probable
arguments, or arguments regarding existence, must be evidently going in a circle, and taking
that for granted, which is the very point in question[1].
Hume ultimately sidesteps this problem by arguing that the intellectual justication
for relying on observations is simply human custom or habit. All inferences from
experience, therefore, are effects of custom, not of reasoning, he writes; and
subsequently: All belief of matter of fact or real existence is derived merely from some
object, present to the memory or senses, and a customary conjunction between that and
some other object[2].
I mention the issue of circularity not because I believe it to be a true shortcoming of
the SM, but rather because it shows how easily many of todays science devotees will
accept the SM based upon an uncritical enthusiasm akin to religious zeal. Surely, one
who bases a belief in the SM solely upon its successful track record is no more
scientic than one who bases a belief in creationism solely upon the Book of Genesis.
Tautology is tautology.
So how does one escape the SMs circularity problem? Chronologically, humanity
had to wait only slightly more than thirty years from the publication of Humes cited
work to a clever resolution provided by the German philosopher Immanuel Kant.
Famously stating that he was roused from his dogmatic slumber by reading Humes
work, Kant spent ten years of his life trying to reconcile the arguments of empiricism
with those of rationalism, a philosophy propounded by writers such as Rene Descartes,
Benedict de Spinoza, and Gottfried Wilhelm von Leibniz. In contrast to empiricism,
rationalism argues that human knowledge can be derived from internal reection
(deduction), and that the experience of the external world is useful primarily for
providing practical examples of internal ideas.
The principal fruit of Kants (1781) ten-year labor was The Critique of Pure Reason,
in which he proposed the philosophy of transcendental idealism. This philosophy
argues that certain types of knowledge, such as mathematical reasoning and the
understanding of time and space, are built into the human mind. Subject to these a
priori conditions, human beings learn about the world through observations as they
are processed by, and conform to, the a priori conditions. Using this type of approach, it
is a simple matter to remove the circularity of the SM by arguing that acceptance of the
SM is just part of the minds a priori construction for example, one could argue that
JRF
8,4
326
we innately believe that a principle that holds true at one particular point in time and
space also should hold true in neighboring points (i.e. during later repetitions of the
same experiment). We then are free to employ the SM to our hearts content.
Feckless frequentism v. bespoke Bayesianism
My primary concern with the SMs philosophical framework is its conventional
formulation in terms of frequentist hypothesis testing. The decision to select the
frequentist approach over its Bayesian counterpart is generally made implicitly, and
for the unstated reason that, since the SM is supposed to represent a collective
decision-making process for all of humanity, it would be impermissible for different
individuals to employ their own individual prior distributions over the unknown
parameters. This idea was expressed clearly by Ronald A. Fisher (1956) in his book
Statistical Methods and Scientic Inference:
As workers in Science we aim, in fact, at methods of inference which shall be equally
convincing to all freely reasoning minds, entirely independently of any intentions that might
be furthered by utilizing the knowledge inferred[3].
While this approach seems quite reasonable in certain situations for example, the
measurement of particle masses in physics or melting points in chemistry there are
other contexts such as nancial planning and decision making in which such a
process is both slow and inefcient.
Consider, for example, the two hurricane-prediction models discussed in a recent
editorial (see Powers, 2006), and suppose (rather fantastically, of course) that the simple
technical model proposed by the author represents the current state of human
knowledge, whereas the more sophisticated fundamental analysis of William Gray
represents a brand new and purportedly superior alternative. Then, under the SM, the
null hypothesis would be given by HO ; Dr Powers model is best for predicting the
number of Atlantic Basin hurricanes in a given year, and the alternative hypothesis
would be H1 ; Dr Grays model is best for predicting the number of Atlantic Basin
hurricanes in a given year.
Putting ourselves in the shoes of an independent academic meteorologist who has
just embarked upon the study of hurricane frequency, it is fairly clear how we should
proceed with our work. Having just learned of Dr Grays model, and how it was shown
statistically to be a better historical predictor of the numbers of Atlantic Basin
hurricanes using a given collection of historical data, we would look for ways to
replicate Dr Grays analysis using different sets of data. Perhaps we would seek
alternative measurements of the same input values used by Dr Gray for the Atlantic
Basin, or perhaps we would compare the two models in a more limited geographical
context (e.g., among hurricanes making landfall in the United States). In either case,
after conducting our study we would publish our results, thereby providing more or
less support to Dr Grays approach, and helping to move forward the cause of good
science.
Now, however, suppose that we are not an academic researcher, but rather the
director of a risk-management and emergency-planning agency for a state or
municipality along the United States Gulf Coast. Suppose also that we have been in
this job for a number of years, and have witnessed rst-hand the relative inaccuracy of
Dr Powers forecasts as they were published year after year. Given the serious
Thoughts on the
scientic
method
327
responsibilities of our position, we might nd ourselves torn between wanting to
embrace wholeheartedly Dr Grays model (because of its apparently greater accuracy)
and sticking with Dr Powers model (because it represents the status quo, and as a
public ofcial, we cannot be too capricious in our decision making). As a consequence,
we may decide to continue using Dr Powers forecasts in our planning, but to
incorporate some information from Dr Grays as well; for example, we may use a
weighted average of the two forecasts, or use Dr Grays forecast only if it provides a
more conservative (i.e. larger) predicted frequency.
In a third scenario, suppose that we are a hedge-fund manager interested in
purchasing nancial securities as part of an overall investment strategy, and that we
are considering buying weather derivatives traded by the Chicago Mercantile
Exchange for Atlanta, Georgia. To understand these instruments better, we do a little
research into weather prediction, and soon come across both Dr Powers and Dr Grays
hurricane forecasts. Because of our relative lack of expertise in this area, we contact a
number of reputable meteorologists, and strangely, keep hearing things like: Dr
Powers is just some insurance guy who doesnt know anything about weather
systems, and Frankly, I dont understand how Dr Powers model came to be so well
established. As a result of this additional information, and given that our motive is
solely to do the very best we can on behalf of our funds investors, we decide to ignore
Dr Powers model completely (i.e. to reject H0 in favor of H1).
From the above scenarios, one can see that R. A. Fishers ideal of an SM that is
equally convincing to all freely reasoning minds, entirely independently of any
intentions is largely an impossible dream. While it may work reasonably well for
academic researchers who are able to proceed at a leisurely pace and who are
uncomfortable relying on private (and therefore subjective) information, it becomes
more problematic for those in need of prompt answers to address important
time-sensitive issues, as well as for those with private information who do not see any
need to be constrained by published results.
The question arises, therefore, whether or not it is reasonable to embrace, explicitly,
an alternative SM let us call it a personalized scientic method (PSM) in which
one is free to: advance the current state of knowledge at ones own pace (i.e. to select
levels of signicance with which one is most comfortable, possibly higher than those
used by most academic researchers); and incorporate private information through the
mechanism of Bayesian analysis.
Clearly, such a program of acquiring knowledge would generate a tremendous
amount of chaos if it were to replace the SM. With every researcher on a different
page from his or her colleagues, there would be no general standard for distinguishing
facts from illusion and truth from wishful thinking. Recalling the inconsistency
problems with the SM listed in the previous editorial, we can see that both the apples
to oranges and moving targets problems would be greatly aggravated by a PSM, as
researchers abandoned any pretense of consensus in the selection of experiments to
conduct and allowed their private beliefs to be unduly inuenced by shing for
statistical signicance.
On the other hand, there would be some benets of a PSM. For one thing, it likely
would counteract the inconsistencies of censorship and double standards mentioned
previously, as researchers would be free to incorporate negative results into their
private beliefs and to offset bias in the selection of alphas with their own discretion in
JRF
8,4
328
that regard. And perhaps the best argument for a PSM already has been hinted at:
Since large numbers of researchers like our public ofcial and hedge-fund manager
already use one, why not simply acknowledge it?
Explicitly embracing a PSM does not have to mean abandoning the conventional
SM. Researchers could continue to follow the SM as before, but simply inject more
opinion, speculation, and subjective judgment into their analyses and conclusions. In
fact, this is exactly what professional actuaries do in the eld of insurance, where
limited data often necessitate intellectual compromise by the averaging of forecasts
and estimates from different sources[4]. Interestingly, insurance actuaries were some of
the earliest researchers to employ Bayesian analyses in their work, and it is
undoubtedly this Bayesian inuence that has allowed them to recognize formally the
usefulness of private (subjective) information. In a world with an accepted PSM,
researchers who found the new approach offensive naturally would be free to express
their opinions, and to organize scholarly journals that adhere to the strictest standards
of the SM (which, as we know, is no mean feat).
One of the supreme ironies of modern intellectual life is the great chasm that exists
between rigorous science, on the one hand, and established religion, on the other. In
todays world, the two opposing camps have reached an uneasy truce under which they
ofcially profess to respect each others domains, and even allow safe passage for some
of their respective adherents to travel back and forth across the bridge that separates
them. However, anyone who tarries on the bridge too long or, God (Darwin?) forbid!
jumps off, is immediately branded a quack or crackpot. A PSM would permit
people to explore this middle ground without fear of ridicule.
So what are we waiting for?
Notes
1. See Hume (1748), Section IV, Part II.
2. See Hume (1748), Section V, Part I.
3. See Fisher (1956), Chapter IV.
4. These techniques are often called credibility methods, because weights are assigned to the
different sources of information based upon their relative credibilities, or intrinsic levels of
believability to the decision maker. This terminology reects a subjective Bayesian
orientation.
References
Fisher, R.A. (1956), Statistical Methods and Scientic Inference.
Hume, D. (1748), An Enquiry Concerning Human Understanding.
Kant, I. (1781), The Critique of Pure Reason.
Powers, M.R. (2006), Catastrophe forecasting: seeing Gray among the black boxes, Journal of
Risk Finance, Vol. 7 No. 5.
Powers, M.R. (2007), Thoughts on the scientic method: part 1 ignorance through
inconsistency, Journal of Risk Finance, Vol. 8 No. 3.
Thoughts on the
scientic
method
329
To purchase reprints of this article please e-mail: reprints@emeraldinsight.com
Or visit our web site for further details: www.emeraldinsight.com/reprints
Calibrating asset correlation for
Indian corporate exposures
Implications for regulatory capital
Arindam Bandyopadhyay, Tasneem Chherawala and Asish Saha
National Institute of Bank Management (NIBM), Pune, India
Abstract
Purpose This paper is a rst attempt to empirically calibrate the default and asset correlation for
large companies in India and elaborate its implications for credit risk capital estimation for a bank.
Design/methodology/approach The authors estimate default probabilities and default
correlations of long-term bonds of 542 Indian corporates using rating transitions and pair-wise
migrations over ten year cohorts of rms. Further, the implicit asset correlation from the estimated
default correlations and default thresholds are derived using the asymptotic single risk factor approach.
Findings The authors nd evidence that default correlations are time variant and vary across
rating grades and industries. The highest correlations are observed between companies within the
same rating grades (systematic risk impact) and within the same industry (industry specic impact).
More interestingly, signicantly smooth monotonic relationship between the probability of default
(PD) and asset correlation as prescribed by the Basel II IRB document (2006) are not found. Moreover,
it is found that the asset correlation range for Indian corporates do not match with what is prescribed
for corporate exposures by BCBS.
Originality/value The authors address the dilemma implied by the negative relationship between
PD and asset correlation as suggested by BCBS IRB formula and other research for developed
economies with estimates of asset correlation for and emerging market like India and demonstrate its
implications on the estimation of credit risk capital.
Keywords Credit, Credit rating, Default, Capital, India, Risk analysis
Paper type Research paper
Introduction
Managing concentration risk in credit portfolios is an integral part of credit risk
management by Banks and FIs. Credit concentration risk may be of two types:
(1) Exposure concentration risk. That is concentration of credit exposure amounts
to a single borrower, or a single group, industry or sector.
(2) Correlation risk. That is, concentration based on common or correlated risk
factors between different borrowers, industries or sectors which may lead to
simultaneous default.
Correlation exists between rms in the same industry because of industry specic
economic conditions. Correlation also exists between companies in different industries
that rely on the same production inputs and among companies that rely on the same
geographical market. Correlation may also arise due to systematic or macroeconomic
conditions in the national or international economy.
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1526-5943.htm
The authors thank an anonymous referee of this journal for extremely helpful suggestions that
helped to improve this paper. They also acknowledge helpful comments from Sanjay Basu.
JRF
8,4
330
The Journal of Risk Finance
Vol. 8 No. 4, 2007
pp. 330-348
qEmerald Group Publishing Limited
1526-5943
DOI 10.1108/15265940710777298
Modeling credit quality correlation and default correlation is therefore crucial for
banks to measure and manage portfolio credit risk. This would require studying the
risk prole of the banks entire credit portfolio and developing the appropriate
methodology for the estimation of default dependence.
Measurement and management of correlation risk in the credit portfolios of banks
has also become an important area of concern for bank regulators worldwide. Indeed,
the BCBS (2006) has specically included an asset correlation factor in the computation
of credit risk capital requirement by banks adopting the Internal Ratings Based (IRB)
approach.
Development of analytical and empirical models for measuring default dependence
by banks is at an embryonic stage in India. Studying default correlation would require
relevant data over a sufciently long time period to capture the common macro
economic factors across economic cycles. However, most Indian banks do not have
organized and consistent historical loan rating data for more than two or three years.
Thus, the major hurdle in developing an analytical model using actual historical loan
default data is the scarcity of data itself. In our study, we therefore consider long term
corporate bond ratings and default information published by an external rating agency
in India as a proxy for rating of long term loans extended by banks to such corporates.
Our study is the rst attempt to empirically derive default correlation and asset
correlation for large companies in India. We estimate:
.
the average transition and default probabilities for various rating grades and
industries;
.
the average default correlations across rating classes, across different bond
grades (Investment vis-a` -vis Non-Investment Grades), and across various
industries; and
.
asset correlations across rating classes and across different bond grades.
Our analysis not only supports the fact that default events are correlated but also
provides numerical estimates of these correlations across rating classes and across
industries for an emerging economy like India. The default correlation values reported
in our study would be key inputs for estimating portfolio credit risk, economic capital
and risk adjusted returns on economic capital for large corporate exposures of banks.
The second important result of our paper is that default correlation between rms
within each rating grade increases as the rating grade worsens, that is, default
correlation is positively related to the default probability (PD) of rms. This result is of
signicance for Indian banks since it implies that poor credit quality commercial loan
portfolios would have to be supported by a higher level of economic capital not just
because the default probabilities in such portfolios would be high but also because of
higher inherent default correlations between poor credit quality borrowers.
The Reserve Bank of India (RBI, 2007) has recently (March 20, 2007) released
revised draft guidelines for the implementation of Basel II norms by banks in India[1].
To facilitate the process of estimating credit risk capital requirements as per the IRB
approach, RBI has suggested that Indian banks should develop their internal rating
methodology, create the default history and estimate default probabilities (PD) and
Loss Given Default (LGD). Two crucial elements that need to be calibrated by the RBI
before the IRB capital computation formula can be formally prescribed for Indian
banks are: the appropriate asset correlation values; and the relationship between PD
and asset correlation in the Indian context.
Calibrating asset
correlation
331
To calibrate the appropriate asset correlation values for the regulatory IRB capital
requirement formula as prescribed by BCBS (2006), we derive asset correlations from our
empirically derived default correlations using the asymptotic single risk factor model. Our
results show that across IG NIG grades, asset correlation decreases with increasing PD,
which is in line with Basel II IRB specications. However, the decrease in asset correlation
with increase in PDacross grades is not smoothly monotonic as it is assumed in the Basel
II The asset correlation range is also different in comparison to what is prescribed for
corporate exposures by BCBS (0.12 minimum to 0.24 maximum). As far as within rating
grades are concerned, we again do not observe any signicant difference in asset
correlation as credit quality deteriorates except between AA-AA v. A-A rating categories.
Here also, the asset correlation range is quite different from the BCBS prescription. This
correlation comparison makes sense because asset correlation would be the key inputs to
constructing the IRB regulatory capital risk weights for banks in India.
The rest of the paper is organized as follows. In the following section we review the
existing literature on default correlations. In the next section, we describe the data, the
variables used in our empirical analysis and the methodology for estimating default
correlation and corresponding implicit asset correlation. In the subsequent sections, we
discuss the major empirical ndings of our paper. The nal section highlights the main
conclusions.
Literature review
Internationally, several methodologies have been developed to estimate default
correlation and several authors have documented the relationship between the initial
credit quality of the portfolio and the default and asset correlation for commercial
portfolios. The structural model approach uses equity correlation as a proxy for asset
correlation. This approach is based on the seminal work by Merton (1974), according to
which loan default occurs when the market value of the rms assets falls below the
book value of debt. Thus, the default correlation between two borrowers is constructed
with the use of correlation of the borrowers asset returns (derived from equity returns)
and the normal inverse of their distance to default. Although intuitive, the said
approach has not been supported by strong empirical analysis that establishes the link
between equity correlation and default correlation. De Servigny and Renault (2002) use
a sample of 1,101 rms from S&Ps 12 industry categories and calculate average equity
correlations across and within industries over the period 1980-2001. They nd that
equity correlation is a very noisy indicator of default correlation. In a recent paper,
Lopez (2004) has used the structural model framework to empirically derive asset
correlations for portfolios of USA, Japanese and European rms. His paper
demonstrates that asset correlation for relatively highly rated, large sized companies
is high. According to his explanation, this relationship arises because high credit
quality rms are more likely to be inuenced by common macro economic conditions.
On the other hand, asset correlations for poor credit quality, large sized companies are
low because defaults of such rms are subject to rm-specic problems. This is also
the relationship which is highlighted by BCBS (2006) wherein, in the credit risk capital
estimation formula for large corporate exposures, asset correlation is a decreasing
function of probability of default.
Zhou (1997) uses the rst-passage-time model within the structural model
framework to derive an analytical formula for calculating default correlations. His
analytical framework is substantiated by empirical results using Moodys default
JRF
8,4
332
database. He concludes that default correlations and underlying asset correlations
have the same sign; the higher the underlying asset correlation, ceteris paribus, the
higher is the default correlation; generally default correlation is lower than the asset
correlation and, the high credit quality of rms (proxied by the leverage) not only
generates a low default probability of each rm but also implies a low default
correlation between rms. Further, high credit quality rms take a longer time to reach
the peak default correlation.
Das et al. (2002) use Moodys database of USA public non-nancial rms to
empirically derive the default correlation structure using the framework of reduced
form, hazard rate models in which the probability of default is determined by
exogenously specied instantaneous default intensity. Their ndings partially
complement those of Zhou (1997) in that they too nd that default probabilities of
issuers are positively correlated. They further observe that default correlations
increase with the level of default risk in the economy, implying a business cycle effect
on default correlations. However, as opposed to Zhou (1997), they conclude that the
highest credit quality issuers have higher default correlation than medium grade rms.
The third approach, used extensively to approximate default correlation, is to
estimate the bivariate transition probabilities and default correlations from actual
historical default data. Lucas (1995), Nagpal and Bahar (2001) and De Servigny and
Renault (2002, 2003) extract information about the joint behavior of rating migrations
and defaults directly from historical bond data to calculate joint default probabilities.
They note that historical rates of default support the idea that credit events are
correlated. However, contradictory to the ndings of Das et al. (2002), they observe that
default correlation increases with the credit risk of rms.
One can clearly see that there are two opposing lines of arguments on the
relationship between PD and default correlation as highlighted above. This therefore
implies that any prior assumption regarding such a relationship for any country would
result in a misspecication of the capital requirement to cushion risk inherent in a
credit portfolio. In the context of these opposing view points, we have made an attempt
to investigate the actual relationship between PD and default correlations (and the
implied asset correlation) for corporates in India.
Data description and methodology for the study
The results presented in our empirical paper are based on CRISILs (a leading credit
rating agency in India now owned by Standard & Poor) annual ratings of long-term
bonds issued by 542 companies from July 1992 to January 2005[2]. Our data base
comprises of large corporates (those which have issued long term bonds in the debt
market). The data on defaulted bonds are, however, available only since January 1995
and hence has restricted the time period of our study. Our analysis focuses on the
default correlations across all rating categories as well as across different industries.
Accordingly, we have studied the rating movements of corporate bonds from
1995-1996 to 2004-2005. The original CRISIL rating scale is divided into fteen
categories (AAA, AA - , . . . , C). We have, however, used seven rating grades (AAA,
AA, A, BBB, BB, B and C) for our analysis. Notches ( - and 2 ) have been
combined into the main rating grades (e.g. all companies with rating grade AA - or
AA 2 have been assigned the grade AA) to ensure adequate sample size of rms in
each grade. The grade D represents Default which is dened as a credit event where
the underlying rm has missed payments (a single days delay or a short fall of even a
Calibrating asset
correlation
333
single rupee in terms of the promised repayment schedule) on a rated instrument. Any
post default recovery is not factored in by CRISILs ratings.
In order to ensure adequate sample size of rms in each industry category, we have
divided the CRISIL corporate bond rating data into 11 industry groups based on their
major economic activities. In classifying rms into different industry categories, we
have harmonized the Prowess CMIE industry categories with NIC 2 digit industry
categories (Table I)[3]. In calculating within industry default correlations, instead of
using seven grades, we have chosen to divide the rated universe into two
categories-investment grade (IG) and non investment grade (IG). Investment grade
ratings correspond to ratings from AAA to BBB, while non investment grade ratings
(NIG) are from BB to C. Again, this has been done to ensure adequate sample size for
each rating grade in each industry.
Our methodology for determining default correlations is based on the approach
developed by Lucas (1995), Nagpal and Bahar (2001) and De Servigny and Renault (2002,
Industry name Industry type
No. of
rms
Auto/parts Manufacture of transport equipments and parts: ships and boats,
railway, commercial vehicles, passenger cars, automobile ancillaries,
two and three wheelers, bicycles, aircrafts 37
Chemical Manufacture of organic and inorganic chemicals and chemical
products, paints, dyes, photographic goods, rubber, plastic, tubes,
petroleum and coal products 116
Diverse Diversied 16
Food products/sugar/
tea/tobacco/beverages
Manufacture of dairy products, sugar, tea. Coffee, vegetable oils and
fats, bakery and food products, beverages, breweries, tobacco and
related products 21
Machine/electrical/
computers
Manufacture of machinery and equipments other than transport
equipments: electronics, electrical, computers, engineering, insulated
wires and cables, re protection equipments, industrial machinery
for food and textile industries, construction 76
Metal/non-metal Basic metal and alloys industries: iron and steel, Ferro alloys,
aluminum, casting of metals, copper, steel tubes, transmission
towers, non metallic mineral products like cement, mica stone, glass
and glass products, ceramic and refractory etc. 65
Other manufacturing Other manufacturing includes optical goods, trade in electrical
machinery, trade in beverages and tobacco, plywood, trade in
minerals and energy sources, trade in textiles, shoe uppers etc. 16
Paper Manufacture of paper and paper products, newsprint and printing,
publishing and allied 15
Power Power generation and electricity generation and transmission 15
Service Hotel, banking, insurance and nancial services 126
Textile Manufacture of cotton textiles, wool, silk and man-made ber
textiles, jute and textile products 39
Total 542
Table I.
Industry categories of
sample rms
JRF
8,4
334
2003). As a rst step, we have done mortality rate analysis of ten one-year cohorts of
companies to nd the number of rms in each rating class in each cohort moving
towards default category (D). Each cohort comprises of all the companies which have a
rating outstanding at the start of the cohort year. From these cohorts, we calculate
year-wise default probabilities for different rating grades and for different industries.
Say there are T
i,D
number of rms migrating to Default category out of N
i
number of
rms in the ith rating grade (or industry) over a one year period, where the subscript i
represents the rating grade (or industry) at the start of the period and the subscript D
represents Default. The one-year probability (PD) of the ith rating grade (or industry) is
estimated by counting the frequencies: T
i;D
=N
i
. The average one-year default
probability for the ith rating grade or industry (PD
i
) is obtained by weighted average,
where the weights are the number of rms in the ith rating class (or industry) in a
particular year divided by the total number of rms in all the years:
PD
i
=
X
n
t=1
w
t
i
T
t
i;D
N
t
i
(1)
where w
t
i
is the weight representing the relative importance of a given year:
w
t
i
=
N
t
i
X
n
s=1
N
s
i
: (1a)
In the second step, we compare the pairs of defaulting rms at the end of the period
with the total number of pairs of rms at the start of the period to count the joint
default frequency for each year from our data. Here we rst consider the joint
migration of two obligors from the same rating grade (say AAA rating) to default
D. Say there are T
i,D
number of rms migrating to Default category out of N
i
number of rms in the ith rating grade, where the subscript i represents the rating
grade at the start of the period and the subscript D represents Default. The one-year
within-grade joint default probability would be: (T
i;D
)
2
=(N
i
)
2
. If obligors are moving
from different rating grades (say AA and B) towards default (D), we should be
interested in pairs of number of rms T
i,D
and T
j,D
migrating respectively to
category D. The one-year between grade joint default probability (JDP) in such a
case would be: T
i;D
T
j;D
=N
i
N
j
. The one-year joint default probabilities are arrived at
by assuming that the defaulting pair of bonds is drawn with replacement. This
assumption is important to ensure that we avoid spurious negative default
correlations in cases where only one default event is observed for a particular rating
category (or industry).
The next step is to calculate the average joint default probabilities (JDP) over time.
For joint default migrations of rms within the same starting rating grade, the
following equation is used:
JDP
i;i
=
X
n
t=1
w
t
i
T
t
i;D

2
N
t
i

2
(2)
where w
t
i
is the weight representing the relative importance of a given year:
Calibrating asset
correlation
335
w
t
i
=
N
t
i
X
n
s=1
N
s
i
: (2a)
For joint default migration of rms from different starting rating grades, we use the
following equation:
JDP
i;j
=
X
n
t=1
w
t
ij
T
t
i;D
T
t
j;D
N
t
i
N
t
j
: (3)
Where, the weight is given as:
w
t
ij
=
N
t
i
- N
t
j
X
n
s=1
(N
s
i
- N
s
j
)
: (3a)
Once we obtain the estimates of average joint default probabilities (JDP) and average
default probabilities (PD), we can calculate the one-year expected default correlations
using the following formula:
r
D;D
i;j
=
JDP
i;j
2PD
i
PD
j

PD
i
(1 2PD
i
)PD
j
(1 2PD
j
)
p : (4)
The above formula (equation 4) is based on the assumption that there is a constant
probability of default (as given by the expected default probability) for a rm or an
industry over time and a constant joint default probability (as given by the average joint
default probability) over time. However, this does not imply that default probabilities or
joint default probabilities are time invariant. It is the limitation of this method.
Empirical results
Business cycle effects on rating transitions
To investigate uctuations of rating transitions over time, we report the Inverse CR
(Inv-CR) in Table II. Inv-CR is the ratio of number of downgrades to number of
upgrades estimated for each year. The rating revisions are visually better captured by
the bars shown in Figure 1 that depict the number of companies whose long-term debt
instruments were upgraded or downgraded between 1995-1996 and 2004-2005. It can
be seen that Inverse CR ratio is increasing from 1995-1996 to 1998-1999 implying that
the number of downgrades to number of upgrades were increasing during this period.
This coincides with a recessionary period in the Indian economy during which there
were a larger number of downgrades than upgrades. The lower Inv-CR during
2000-2001, 2003-2004 and 2004-2005 reects a period of relative macro economic
stability of the Indian economy (Government of India, 2004-2005).
Business cycle effects on default probabilities
In Tables III to V, we analyze the rating transition probabilities and default
probabilities for various rating grades. In Table III, we present a one-year average
transition matrix for the entire sample period (1995-1996 to 2004-2005). We nd that as
JRF
8,4
336
the credit quality worsens (i.e. rating grades decline) the default probability (PD)
increases. One can see from said table that PD jumps sharply from 5.17 per cent for
BBB to 28.93 per cent for BB. This justies our classication of bonds in BB grade and
below as Non Investment Grade (NIG) and AAA to BBB grades as Investment Grade
(IG)[4]. We also nd that the rating stability declines as the credit quality worsens[5]. It
is worthwhile to mention that our produced transition matrix also looks similar to
CRISIL except the data sample since we are studying the performance of a balanced
sample of corporates over the eleven years. CRISIL has not done any study on default
correlation. They also do not report industry wise PDs. Interestingly, ours or CRISILs
transition matrix and PDs across rating grades is markedly different from Standard
and Poors transition matrix specially for non investment grades (NIG) companies. One
compare that our below BBB rms are as risky as S&P CCC rms are.
Year Inv-CR Upgrades Downgrades
1995-1996 1.09 11 12
1996-1997 4.22 9 38
1997-1998 18.00 4 72
1998-1999 19.00 5 95
1999-2000 9.40 5 47
2000-2001 1.33 18 24
2001-2002 20.00 1 20
2002-2003 6.00 3 18
2003-2004 0.56 9 5
2004-2005 0.22 9 2
Table II.
Downgrades v. upgrades
Figure 1.
Calibrating asset
correlation
337
In Table IV and Table V we present two average one-year transition matrices for
the periods 1995-1996 to 1999-2000 and 2000-2001 to 2004-2005. Except for the
rating grade BBB and C, the one year PDs for all rating grades are higher for the
rst period as compared to the second period. This is consistent with the results of
the Inverse CR analysis. This result indicates that default probabilities vary over
time and on an average, PDs are higher during the recessionary period of the Indian
Economy.
Year 2
Percentages
Year 1 AAA AA A BBB BB B C D
AAA 97.08 2.92 0.00 0.00 0.00 0.00 0.00 0.00
AA 2.54 87.57 7.93 1.05 0.60 0.15 0.00 0.15
A 0.00 4.35 79.97 9.14 3.48 0.44 0.73 1.89
BBB 0.00 0.74 5.90 67.53 14.76 2.21 3.69 5.17
BB 0.00 0.83 0.00 1.65 57.02 4.13 7.44 28.93
B 0.00 0.00 0.00 7.41 0.00 55.56 7.41 29.63
C 0.00 0.00 0.00 2.33 0.00 0.00 51.16 46.51
D 0.00 0.00 0.31 0.31 0.92 0.00 0.00 98.46
Table III.
One-year average
transition matrix
(1995-1996 to 2004-2005)
Year 2
Percentages
Year 1 AAA AA A BBB BB B C D
AAA 97.50 2.50 0.00 0.00 0.00 0.00 0.00 0.00
AA 3.34 92.10 3.95 0.61 0.00 0.00 0.00 0.00
A 0.00 9.46 80.41 6.08 0.68 0.00 2.03 1.35
BBB 0.00 1.69 8.47 69.49 6.78 5.08 0.00 8.47
BB 0.00 3.23 0.00 6.45 74.19 3.23 0.00 12.90
B 0.00 0.00 0.00 13.33 0.00 53.33 6.67 26.67
C 0.00 0.00 0.00 4.35 0.00 0.00 47.83 47.83
D 0.00 0.00 0.41 0.41 1.22 0.00 0.00 97.97
Table V.
One-year average
transition matrix
(2000-2001 to 2004-2005)
Year 2
Percentages
Year 1 AAA AA A BBB BB B C D
AAA 96.12 3.88 0.00 0.00 0.00 0.00 0.00 0.00
AA 1.77 83.19 11.80 1.47 1.18 0.29 0.00 0.29
A 0.00 3.48 93.91 11.74 5.00 0.65 0.43 2.39
BBB 0.00 0.47 5.19 66.98 16.98 1.42 4.72 4.25
BB 0.00 0.00 0.00 0.00 51.11 4.44 10.00 34.44
B 0.00 0.00 0.00 0.00 0.00 58.33 8.33 33.33
C 0.00 0.00 0.00 0.00 0.00 0.00 55.00 45.00
D 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00
Table IV.
One-year average
transition matrix
(1995-1996 to 1999-2000)
JRF
8,4
338
Business cycle effects on joint default frequencies
In Figure 2, we plot the joint default frequency over the period of ten years between
IG-IG, NIG-NIG and IG-IG bonds. It is quite clear from these diagrams that joint default
frequencies within the NIG group have a more pronounced co-movement with the
economic cycle. The year 1998-1999 has witnessed a slight increase in correlated default
between the IG-NIG grades and even within the IG group. The joint default movements,
in general, are lower during the economic recovery period (2002-2003 to 2003-2004).
It is clear from the ndings in the above three sections that rating transitions,
default probabilities and joint default frequencies vary over the business cycle. This
clearly indicates that default correlations are time varying and would tend to increase
during the recessionary phase. Further, the business cycle impact on the default
correlation would be higher for the low credit quality (NIG-NIG) rms as compared to
the IG-IG group or IG-NIG group.
Industry wise default probabilities
Table VI presents the average default probabilities for each of the 11 industries. As
would be expected, the PD for each industry is higher for the NIG class as compared to
the IG class. The PD for the NIG class as a whole (all industries) is 3.30 per cent and the
PD for the IG class as a whole is 1.42 per cent. The highest default probabilities are
observed for Machine/Electrical/Computers industry (7.51 per cent) followed by Food
Products industry (7.41 per cent)
Default correlation across rating grades
Table VII reports the one-year average joint default probabilities across rating classes.
To derive the numerical estimates of default correlation and to have a better
understanding of the relationship between default correlation and the default risk of
borrowers, we calculate the average default correlation across rating grades using
equation (4). The results are presented in Table VIII.
In Table VIII, we report the estimates of default correlation across rating classes.
We have also tested whether the correlation coefcients are statistically different from
zero for validating these gures. The signicance of the correlation coefcient is
computed by testing the hypothesis that the correlation is zero by using the following
formula:
t =
r

n 22
_

1 2r
2
_
where r is the correlation coefcient and n is sample size, and where one looks up the t
value in a table of the distribution of t, for (n 2 2) degrees of freedom. If the computed t
value is as high or higher than the table t value, then the researcher concludes the
correlation is signicant (that is, signicantly different from 0). The t values and the
degrees of freedom of the default correlation coefcients are reported in the
parentheses below each coefcient in Table VIII.
The rst important conclusion that we draw from Table VIII is with regard to the
estimated values of the default correlations. The values of signicant positive default
correlations are much higher than those quoted for developing economies. On the other
hand, we also report a signicant negative default correlation between medium and
low credit quality grades. For comparison, we show in Tables IX and X, across rating
Calibrating asset
correlation
339
Figure 2.
Joint default frequency
JRF
8,4
340
category default correlation matrices derived by De Servigny and Renault (2004) and
Lucas (1995) respectively.
The second important conclusion that we draw from Table VIII is that with
declining credit quality, default correlation increases within the same rating grade (for
example, the default correlation within AA rating grade is 1.222 per cent and it
increases up to 24.55 per cent within B rating grade). Most of the within grade default
correlations are positive and signicantly different from zero (Table VIII). This table
also shows that there is relatively lower default correlation between the lower grade
categories than between the higher rating grades (for example, default correlation
between A and AA is 1.71 per cent vis-a` -vis default correlation of 0.32 per cent between
B and BBB). Further, default correlation is low and in some cases, negative, between
the low rating grades (B and C) and higher rating grades (A and AA). As far as
negative correlations are concerned, we nd that except for default correlation between
A and B (26.65 per cent), all other negative default correlations are marginal. This
may be a result of low sample size of defaulted bonds in these categories.
Table XI best captures the effect of the common macroeconomic environment on
correlated defaults. The results reect that default correlation within IG group is lower
than the default correlation within the NIG group over a one-year horizon for all
industries taken together. This further reinforces our earlier conclusion that as we
move from IG-IG to NIG-NIG category (i.e. credit quality is worsening and PD is
Percentages
Industry IG NIG ALL
Auto/parts 1.79 16.67 2.87
Chemical 0.88 32.00 3.98
Diverse 3.70 18.75 7.14
Food products/sugar/tea/tobacco/beverages 1.37 55.56 7.41
Machine/electrical/computers 2.75 37.14 7.51
Metal/non-metal 3.11 35.48 6.60
Other manufacturing 0.00 0.00 0.00
Paper 0.00 33.33 5.66
Power 0.00 0.00 0.00
Service 0.22 27.78 1.25
Textile 1.52 40.00 5.44
All 1.42 32.98
Table VI.
Industry-wise average
one- year PD for IG and
NIG and Pool (1995-1996:
2004-2005)
Year 2
Percentages
Year 1 AAA AA A BBB BB B C
AAA 0.000
AA 0.000 0.002
A 0.000 0.012 0.085
BBB 0.000 0.028 0.186 0.655
BB 0.000 0.098 0.737 2.222 11.052
B 0.000 0.000 0.146 1.563 6.751 13.898
C 0.000 0.066 0.689 2.174 11.881 12.088 25.748
Table VII.
One-year average joint
default probabilities
across rating classes
(1995-1996 to 2004-2005)
Calibrating asset
correlation
341
Y
e
a
r
2
A
A
A
B
B
B
B
B
B
C
Y
e
a
r
1
%
t
-
v
a
l
u
e
(
%
)
a
%
t
-
v
a
l
u
e
(
%
)
a
%
t
-
v
a
l
u
e
(
%
)
a
%
t
-
v
a
l
u
e
(
%
)
a
%
t
-
v
a
l
u
e
(
%
)
a
%
t
-
v
a
l
u
e
(
%
)
a
A
A
1
.
2
2
2
0
.
4
5
,
1
3
3
6
A
1
.
7
1
0
.
6
3
,
1
3
5
7
2
.
6
8
1
.
0
0
,
1
3
7
8
B
B
B
2
.
3
8
0
.
7
3
,
9
3
9
2
.
9
3
0
.
7
7
,
6
9
0
7
.
9
3
*
*
1
.
8
5
,
5
4
2
B
B
3
.
1
0
0
.
8
7
,
7
8
9
3
.
1
0
0
.
8
8
,
8
1
0
7
.
2
5
*
1
.
4
4
,
3
9
2
1
3
.
0
6
*
*
2
.
0
4
,
2
4
2
B
2
2
.
5
1
2
0
.
6
6
,
6
9
5
2
6
.
6
5
*
*
2
1
.
7
8
,
7
1
6
0
.
3
2
0
.
0
5
5
,
2
9
8
2
8
.
7
9
2
1
.
0
7
,
1
4
8
2
4
.
5
5
*
*
1
.
8
3
,
5
4
C
2
0
.
1
9
2
0
.
0
5
1
,
7
1
1
2
2
.
7
9
2
0
.
7
5
,
7
3
2
2
2
.
0
8
2
0
.
3
7
,
3
1
4
2
6
.
9
6
2
0
.
8
9
,
1
6
4
2
7
.
4
3
2
0
.
6
1
,
7
0
1
6
.
5
3
7
*
1
.
5
4
,
8
6
N
o
t
e
s
:
a
t
-
v
a
l
u
e
s
a
n
d
t
h
e
n
u
m
b
e
r
o
f
o
b
s
e
r
v
a
t
i
o
n
s
r
e
s
p
e
c
t
i
v
e
l
y
;
*
d
e
n
o
t
e
s
s
i
g
n
i

c
a
n
c
e
a
t
5
-
1
0
%
;
*
*
d
e
n
o
t
e
s
t
h
e
s
i
g
n
i

c
a
n
c
e
l
e
v
e
l
a
t
5
%
o
r
b
e
t
t
e
r
Table VIII.
One-year default
correlations across rating
classes (1995-1996 to
2004-2005)
JRF
8,4
342
increasing), default correlation is increasing. We can explain this by the fact that NIG
grade rms can be considered as a portfolio of poor credit quality rms with different
rating grades including BB, B and C. Such portfolio would be relatively more sensitive
to common risks than to idiosyncratic risks. If the economy experiences a downturn, all
the rms within NIG class are more likely to experience problems leading to defaults
Default correlation across industries
We have applied the methodology described previously to compute default
correlations across industries. The results are shown in Table XII. The highest
correlations can be observed in the diagonal of the correlation matrix, i.e. within the
same industry. This is obvious, since rms belonging to the same industry are more
likely to default simultaneously. Thus, irrespective of growth prospects, risk is high if
the banks portfolio is more concentrated in a single industry. The off-diagonal
Year 2
Percentages
Year 1 AA A BBB BB B CCC
AA 0.16
A 0.02 0.12
BBB 20.03 0.03 0.33
BB 0.00 0.19 0.35 0.94
B 0.10 0.22 0.30 0.84 1.55
CCC 0.06 0.26 0.89 1.45 1.67 8.97
Sources: De Servigny and Renault (2004), Standard & Poors CreditPro
Table IX.
One-year default
correlations all countries,
all industries, 1981-2002
Year 2
Percentages
Year 1 Aaa Aa A Baa Ba B
Aaa 0
Aa 0 0
A 0 0 0
Baa 0 0 0 0
Ba 0 0 0 0 2
B 0 1 0 1 4 7
Note: Moodys Investors Service ratings of all non-municipal issuers
Source: As reported by Lucas (1995)
Table X.
One-year default
correlations, 1970-1993
IG NIG
% t-value (%)
a
% t-value (%)
a
IG 2.08
*
1.31,3942
NIG 1.43 0.66,2162 6.59
*
1.29,382
Notes:
*
Denotes signicance level at 5-10%;
a
t-values and the degrees of freedom respectively
Table XI.
One-year default
correlations across IG
and NIG grades,
1995-1996 to 2004-2005
Calibrating asset
correlation
343
P
e
r
c
e
n
t
a
g
e
s
A
u
t
o
/
p
a
r
t
s
C
h
e
m
i
c
a
l
D
i
v
e
r
s
e
F
o
o
d
/
s
u
g
a
r
/
t
e
a
M
a
c
h
i
n
e
/
e
l
e
c
t
r
i
c
a
l
/
e
l
e
c
t
r
o
n
i
c
s
M
e
t
a
l
/
n
o
n
-
m
e
t
a
l
O
t
h
e
r
m
a
n
u
f
a
c
t
u
r
i
n
g
P
a
p
e
r
P
o
w
e
r
S
e
r
v
i
c
e
T
e
x
t
i
l
e
A
u
t
o
/
p
a
r
t
s
2
.
9
6
C
h
e
m
i
c
a
l
0
.
0
0
2
.
8
1
D
i
v
e
r
s
e
2
0
.
6
1
3
.
1
2
2
1
.
2
6
F
o
o
d
/
s
u
g
a
r
/
t
e
a
2
0
.
6
1
1
.
9
6
1
0
.
7
0
1
0
.
4
4
M
a
c
h
i
n
e
/
e
l
e
c
t
r
i
c
a
l
/
e
l
e
c
t
r
o
n
i
c
s
1
.
0
2
2
.
6
1
1
2
.
3
7
5
.
8
9
9
.
7
7
M
e
t
a
l
/
n
o
n
-
m
e
t
a
l
1
.
3
2
2
.
4
0
4
.
4
0
2
.
8
1
5
.
0
0
4
.
3
8
O
t
h
e
r
m
a
n
u
f
a
c
t
u
r
i
n
g
0
.
0
0
0
.
0
0
0
.
0
0
0
.
0
0
0
.
0
0
0
.
0
0
0
.
0
0
P
a
p
e
r
0
.
0
0
0
.
0
0
0
.
0
0
0
.
0
0
0
.
0
0
0
.
0
0
0
.
0
0
0
.
0
0
P
o
w
e
r
0
.
0
0
0
.
0
0
0
.
0
0
0
.
0
0
0
.
0
0
0
.
0
0
0
.
0
0
0
.
0
0
0
.
0
0
S
e
r
v
i
c
e
1
.
3
2
1
.
0
3
5
.
8
9
3
.
7
2
5
.
4
5
2
.
9
6
0
.
0
0
0
.
0
0
0
.
0
0
4
.
2
8
T
e
x
t
i
l
e
0
.
2
4
1
.
6
2
2
0
.
3
6
1
.
9
6
2
0
.
0
2
0
.
6
3
0
.
0
0
0
.
0
0
0
.
0
0
0
.
3
5
5
.
3
7
Table XII.
One- year default
correlations across
industries, period:
1995-2006 to 2004-2005
JRF
8,4
344
correlations tend to be relatively low. This would give banks ample opportunities to
exploit diversication strategies across industries and thereby reduce the unexpected
losses on their large commercial loan portfolios.
Deriving implied asset correlation from default correlation
In order to compare the results of our study to the Basel II credit risk capital
formulation, we need to derive the implicit asset correlation from our estimates of
default correlation. For this, we have followed the asymptotic single risk factor based
approach. In this approach, the key risk driver is the value of the rms assets. It is
assumed that each rm earns some random return on its assets (A
i
) at the horizon
(T = 1 year). If the return is sufciently negative, the value of the rms assets falls
beneath the value of its xed liabilities, and the rmdefaults. The realized returns are a
weighted sum of a single common risk factor (representing systematic risk) and a
shock idiosyncratic to the rm (Gordy and Heiteld, 2002). Both of these follow
standard normal distribution. The idiosyncratic component is uncorrelated with the
systematic component and with other rms idiosyncratic components.
We assume A
i
, N(0; 1); when A
i
falls below a critical default threshold say K
i
(popularly known as distance to default), default is triggered. Thus:
PD
i
= Pr[A
i
# K
i
]
= K
i
= N
21
(PD
i )
)
: (5)
The K
i
cut-off level is thus a function of the ith rms rating grade.
The joint default probability can therefore be obtained by using the following
expression:
JDP
ij
= Pr[A
i
# K
i
; A
j
# K
j
] = N
2
(K
i
; K
j
; r
a
ij
) (6)
where, N
2
(.) denotes the cumulative bivariate standard normal distribution, and, r
a
ij
denotes the asset correlation between rm i and rm j.
Given the values of JDP
ij
(empirically derived), K
i
and K
j
(from equation 5 above),
we can infer the implied asset correlation (r
a
ij
) by inverting the joint distribution:
r
a
ij
= N
21
2
(K
i
; K
j
; JDP
ij
): (7)
The implied asset correlations within IG (r
a
IG2IG
) and NIG (r
a
NIG2NIG
) grade rms and
across IG-NIG (r
a
IG2NIG
) grade rms are estimated used the methodology described
above, presented in Table XIII.
The asset correlation estimates reveal that as PD increases (credit quality worsens
from IG to NIG grade), asset correlation decreases. This relationship is the opposite of
the relationship between PDand default correlation as observed in Table VIII. We have
also tested whether these variation in the asset correlation observed across grades is
statistically signicant. A t test is used to test for the difference of two correlations
assuming independent samples. For this, we have rst used Fishers z score
transformation technique to convert these correlation estimates into a z-score for
purpose of our hypothesis testing. This is done by dividing the correlation plus 1, by
the same correlation minus 1; then taking the natural log of the absolute value of the
result; then divide that result by 2[6]. Next, we estimate the standard error of difference
Calibrating asset
correlation
345
between the two correlations[7]. After obtaining z scores and standard errors (SE), we
compute the t value by dividing the difference between two z scores by the standard
error. The t values along with the number of observations are reported in the
parentheses of the asset correlation coefcients. One can see that there is signicant
asset correlation difference across IG-IG v. NIG-IG and IG-IG v. NIG-NIG. However,
NIG-IG v. NIG-NIG difference is not signicant. Hence, the decrease in asset correlation
with increase in PD across grades is not smoothly monotonic as assumed in the Basel II
(2006) document.
We also estimate implied asset correlations within each rating grade starting from
AA to C to nd if there is any difference in asset correlation in a more granular way.
The results from Table XIV show that there is no signicant monotonic relationship
between PD and asset correlation within rating grades. In fact, we observe that though
the asset correlation within AA grade is high, the correlation value signicantly falls
for A grade and subsequently increases (though increase is statistically insignicant)
as the credit quality further worsens (up to B grade). This empirically observed
relationship is in contradiction to the Basel II asset correlation formula which assumes
a smooth inverse relationship between PD and asset correlation.
Conclusions
The main aim of this paper is to derive estimates of default correlations and asset
correlations for large companies in India. Our empirical results substantiate the
presence of default correlation due to macroeconomic factors and/or industry-specic
factors. Since default events are not independent, as is evident from our observations,
the correlation effects need to be considered carefully in managing and measuring
concentration risks in credit portfolios. We also nd corroborating evidence that
default correlations vary over time due to the business cycle effect and that default
N t-values
a
IG-IG 0.162
NIG-IG 0.035
*
4.79
NIG-NIG 0.029 20.11
Notes:
a
t-values of the difference between the correlation coefcients; the t value for the difference
between correlation coefcients of IG-IG v. NIG-NIG is 2.5 and is signicant at 1% level;
*
denotes the
signicance level at 5% or better
Table XIII.
Implicit asset correlations
across IG and NIG grades
N t-values
a
AA-AA 0.35
A-A 0.19
*
4.50
BBB-BBB 0.22 20.62
BB-BB 0.29 20.96
B-B 0.33 20.29
C-C 0.28 0.31
Notes:
a
t-values of the difference between the correlation coefcients;
*
denotes the signicance level
at 5% or better
Table XIV.
Implicit asset correlations
within rating grades
JRF
8,4
346
correlations vary across rating grades and industries. The highest default correlations
are observed between companies within the same rating grades (systematic risk
impact) and within the same industry (industry specic impact).
Most interestingly, we do not nd signicantly smooth monotonic relationship
between the probability of default (PD) and asset correlation as prescribed by the Basel
II IRB document (2006). Our results show that across IG NIG grades, asset correlation
decreases with increasing PD, which is in line with Basel II IRB specications.
However, the decrease in asset correlation with increase in PD across grades is not
smoothly monotonic as it is assumed in the Basel II. The asset correlation range is also
different in comparison to what is prescribed for corporate exposures by BCBS (0.12
minimum to 0.24 maximum). As far as within rating grades are concerned, we again do
not observe any signicant difference in asset correlation as credit quality deteriorates
except between AA-AA v. A-A rating categories. Here also, the asset correlation range
is quite different from the BCBS prescription.
These ndings have serious implications for banks which will compare their IRB
regulatory capital with their economic capital as in the case of economic capital there
have more exibility. To further elaborate, this may encourage banks to arbitrage their
regulatory capital by lending to poor credit quality corporates. Moreover, since PDs of
lower grades are higher in comparison to USA ratings, Basel II correlation would be
lower in the non-investment grade categories corporate assets in emerging market (like
India) than the developed market. This will further encourage capital arbitrage
activities. The Basel II correlation function (in which correlation is a declining function
of PD) is quite possibly justied more on political grounds (i.e. a desire to favor small
and medium enterprises) than with reference to empirical evidence.
Notes
1. The Reserve Bank of India has stipulated that Indian banks with overseas presence and
branches of foreign banks in India will adopt Basel II norms with effect from March 31, 2008
and other banks not later than March 31, 2009. Subject to meeting some minimum qualifying
criteria, some banks may be allowed to migrate to IRB Approach.
2. External credit rating agencies in India are: Credit Rating Information Services of India
Limited (CRISIL), Investment Information and Credit Rating Agency (ICRA) and Credit
Analysis and Research Limited (CARE). CRISIL began its operations in 1987. In 1992, the
Reserve Bank of India and the Securities and Exchange Board of India (the main regulators
of Indias nancial markets) made ratings mandatory for various classes of debt
instruments. ICRA was established in the year 1991. CARE was incorporated in April 1993.
3. CMIE and NIC are corporate databases in India. For a discussion on concordance, see
Veeramani (2001), Debroy and Santhanam (1993).
4. In our sample, a few companies which were assigned Default Grade for their long term debt
have subsequently been upgraded to Non-Default Categories, thereby implying that the PD
for D grade is not 100 per cent. Thus the Default grade is not an all absorbing state.
5. Stability of ratings implies that the probability that a company would maintain its rating
grade at the end of the period is higher than its migration to any other rating grade during
the period.
6. The formula for z score: z = ln[}(r - 1)/r 2 1)} ]/2 where r is the default correlation
coefcient.
7. Standard error (SE) =

1
n
1
23
-
1
n
2
23
q
where n
1
and n
2
are the sample sizes of two
independent samples.
Calibrating asset
correlation
347
References
Basel Committee on Banking Supervision (BCBS) (2006), International Convergence of Capital
Measurement and Capital Standards: A Revised Framework, Publication No. 128, Bank for
International Settlements, Basel, June.
Das, S.R., Freed, L., Geng, G. and Kapadia, N. (2002), Correlated default risk, working paper,
Default Risk.
De Servigny, A. and Renault, O. (2002), Default correlation: empirical evidence, working paper,
Standard and Poors.
De Servigny, A. and Renault, O. (2003), Correlations evidence, RISK, July, pp. 90-4.
De Servigny, A. and Renault, O. (2004), Measuring and Managing Credit Risk, Chapter 5,
Standard & Poors, McGraw-Hill Companies, Inc., New York, NY.
Debroy, B. and Santhanam, A.T. (1993), Matching trade codes with industrial codes, Foreign
Trade Bulletin, Indian Institute of Foreign Trade.
Government of India (2004-2005), Economic Survey 2004-2005, Government of India.
Gordy, M. and Heiteld, E. (2002), Estimating default correlations from short panels of credit
rating performance data, working paper, Federal Reserve Board, Washington DC.
Lopez, J.A. (2004), The empirical relationship between average asset correlation, rm
probability of default, and asset size, Journal of Financial Intermediation, Vol. 13 No. 2,
pp. 265-83.
Lucas, D.J. (1995), Default correlation and credit analysis, Journal of Fixed Income, Vol. 4 No. 4,
pp. 76-87.
Merton, R.C. (1974), On the pricing of corporate debt: the risk structure of interest rates, Journal
of Finance, Vol. 29, pp. 449-70.
Nagpal, K. and Bahar, R. (2001), Measuring default correlation, RISK, Vol. 14, March,
pp. 129-32.
Reserve Bank of India (2007), Revised Draft Guidelines for Implementation of the New Capital
Adequacy Framework, March 20, RBI, Mumbai.
Standard & Poors (n.d.), Standard & Poors Credit Por, McGraw-Hill, New York, NY.
Veeramani, C. (2001), Analyzing trade ows and industrial structure of India: the question of
data harmonization, working paper, CDS , Thiruvanthapuram.
Zhou, C. (1997), Default correlation: an analytical result, Mimeo, Federal Reserve Board,
Washington, DC.
Further reading
Blalock, H. (1972), Social Statistics, McGraw-Hill, New York, NY, pp. 406-7.
Corresponding author
Arindam Bandyopadhyay can be contacted at: arindam@nibmindia.org
JRF
8,4
348
To purchase reprints of this article please e-mail: reprints@emeraldinsight.com
Or visit our web site for further details: www.emeraldinsight.com/reprints
Dividend policy and payout ratio:
evidence from the Kuala Lumpur
stock exchange
Abdulrahman Ali Al-Twaijry
Accounting Department, College of Business and Economics,
Qassim University, Qaseem, Saudi Arabia
Abstract
Purpose The purpose of this research is to identify the variables with an expected inuence on
dividend policy and on payout ratio in an emerging market.
Design/methodology/approach Based on the literature, eight hypotheses were developed and
tested using 300 rms randomly selected from the Kuala Lumpur Stock Exchange. Additional
statistical analyses were presented.
Findings The results suggest that current dividends are affected by their pasts and their future
prospects. To a lesser extent dividends were associated with net earnings. Payout ratios (POR) were
not found to have a strong effect on the companys future earning growth, but had some signicant
negative correlation with the companys leverage. Cash per share and share book value signicantly
and positively affect both DPS and POR.
Practical implications The ndings of the study might be of interest to academicians and
practitioners.
Originality/value This paper explores the dividend policy and the payout ratio of listed
companies in a fast-growing market that has received inadequate research attention. The paper thus
adds to the body of accounting knowledge.
Keywords Dividends, Stock exchanges, Malaysia, Earnings per share
Paper type Research paper
Introduction
The dividends and dividend policy were the subject of many studies for many years
from past to present (e.g. Lintner, 1956; Gordon, 1959; Miller and Modigliani, 1961;
Mancinelli and Ozkan, 2006; Amidu and Abor, 2006; Zhou and Ruland, 2006). Since
dividends have an effect on stock prices and companys future growth, corporate
governance should have a suitable dividend strategy. There are several dividend
approaches such as stable dividends, payout ratio, and cash availability. Corporate
management needs to take different variables into account before taking the decision
on the how and the when of dividend payout. Researchers follow different procedures
(theoretical and/or empirical), endeavoring to highlight factors expected to have some
inuence on dividend payout decisions and policies. However, most of this empirical
work has been focused on companies listed in markets of developed nations. Therefore
the conclusions reached may not be applicable in countries with different corporate
cultures and economic frameworks.
In this study, we thus examine the impact of various factors on dividend policy and
payout ratios in the semi-developed market of Malaysia. This country was chosen
because it represents a different market with different features. It is a fast-growing
market in a state between developed and developing. Thanoon et al. (2005) stated,
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1526-5943.htm
Dividend policy
and payout ratio
349
The Journal of Risk Finance
Vol. 8 No. 4, 2007
pp. 349-363
qEmerald Group Publishing Limited
1526-5943
DOI 10.1108/15265940710777306
Malaysia [is] setting out to transform its economy through industrialization to become
a fully developed country by the year 2020 (VISION, 2020) . . . This form of economy
might require a special dividend policy that may differ from those used in developed or
developing markets. Based on a sample of 300 Malaysian listed companies on the
Kuala Lumpur Stock Exchange, several factors including Net EPS, cash available PS,
book value of the share, company size, company age, past dividends, and past and
future earnings will be discussed. Correlation and mean comparison analyses were
employed to examine the possible effect of these factors.
Literature review and hypotheses development
Dividend policy and dividend payout ratios have occasioned a large volume of research
and are still attracting researchers. The great majority of the studies on this subject,
however, were based on developed markets, especially those of the USA. and Europe,
and little concern was given to less developed nations. Some studies, for example
Lintner (1956), Baker et al. (1985), Pruitt and Gitman (1991), Benartzi et al. (1997), Baker
and Powell (2000), investigated the possible effect of past dividends on future earnings
and/or dividends. Some other researchers focused on the effect of investment decisions
of rms (Fama, 1974), industry classication (Baker, 1988), capital adequacy (Dickens
et al., 2002), and the ownership structure of companies (Mancinelli and Ozkan, 2006) on
dividend policy. Baker et al. (2001) based their research on a survey of NASDAQ-listed
rms to test twenty-two different factors that might inuence the dividend policy. The
analysis of the survey questionnaires showed that the most important determinants of
dividend strategies are:
.
the pattern of past dividends;
.
stability of earnings; and
.
current and expected future earnings.
Fama and French (2001) were more concern with disappearing dividends and the
disappearances real cause (be it the change of a rms characteristics or a lower
propensity to pay). They reported that the percentage of USA publicly held companies
paying dividends declined from 66.5 per cent in 1978 to 20.8 per cent in 1999, and that
this decline was partially caused by the changing characteristics of rms and was
partially due to just less desire. The ndings of DeAngelo et al. (2004) contradicted
those of Fama and French since their results, which were based on aggregate dividends
from 1978, showed an increase in dividends in recent years. It should be noticed that
using aggregate data can sometimes be misleading when examining the corporate
dividend trend because very large companies are more likely to dominate the
aggregate results.
Arnott and Asness (2003) have challenged the familiar wisdom. Such wisdom
advocates that a higher payout ratio results in low future growth. Arnott and Asness
based their study on America stock market (S&P 500) and found that higher
aggregate-dividend-payout ratios were associated with higher future earning growth.
Both Zhou and Ruland (2006) and Gwilym et al. (2006) supported the ndings of Arnott
and Asness. Zhou and Ruland examined the possible impact of dividend payouts on
future earning growth. Their study used a sample of active and inactive stocks listed
on NYSE and NASDAQ with positive, non zero, payout ratio companies covering the
period from 1950 to 2003. The ndings showed that the payout ratio mean was 0.40
JRF
8,4
350
while the median was 0.33. Also their reported regression results showed a strong
positive relation between the payout ratio and the future earnings growth. Gwilymet al.
(2006) have extended the work of Arnott and Asness to include eleven international
markets, the majority of which were European. In addition to their support of Arnott
and Asness ndings, Gwilym et al. (2006) found that higher payout ratios do not lead
to higher, real dividend growth. From a different angle, Mancinelli and Ozkan (2006)
undertook an empirical investigation of the relationship between the ownership
structure of companies and dividend policy using 139 rms listed in Italian exchange.
Their results suggested that the payout ratio is negatively associated with the voting
rights of the largest shareholders
These studies covered the USA and Europe markets of mainly developed nations,
but a few studies endeavored to investigate dividend policy and related factors in other
markets. For example, Singhania (2005), Amidu and Abor (2006) studied the dividend
payout in less developed countries. Singhania studied the historical trends of dividend
payout using a sample of Indian companies from Bombay Stock Exchange. He found
out that companies paying dividends dropped from 448 in 1992 to 376 in 2004, but that
the mean of the payout ratio increased with a trend volatility of between 25 per cent
and 68 per cent during the examined period, and more than 50 per cent of the studied
companies had been following a dividend strategy that had part of the prot
distributed to shareholders and the rest for retention. Amidu and Abor examined the
factors affecting dividend payout ratios of listed companies in Ghana. The results of
their study showed that payout ratios were positively correlated to protability, cash
ow and tax but were negatively related to risk and growth.
Based on this literature, eight null hypotheses can be advanced:
H1. There is no relationship between current and past or future dividends.
H2. The dividends are not affected by current, past or future earnings.
H3. Firm-specic factors (size, age and sector) have no impact of on the amount of
dividends paid out.
H4. There is no relationship between current payout ratio and current, past or
future growth.
H5. There is no relationship between the current payout ratio and its past and
future.
H6. Payout ratios do not have an effect on corporate future earnings.
H7. The corporate leverage has no negative impact on the payout ratio.
H8. The payout ratio has no signicant impact on the companys future earnings
growth
Data collection and descriptive analysis
The sample of this study is cross-sectional and consists of 300 Malaysian companies
randomly selected from the Kuala Lumpur Stock Exchange (KLSE) listed in the
Datastream electronic companies database (total 646 rms). The nancial
accounting data were collected for ve years (from 2001 to 2005) in addition to other
important company features such as foundation date and sector. After the data were
Dividend policy
and payout ratio
351
retrieved into Excel, they were checked and corrected for any misspecication and
order. Tables I, II and III) disclose the characteristics of the selected companies.
Only a few companies (5 per cent) are of the small size with total assets less than 100
million Malaysian Ringgit (MYR). But half the companies are of the mid-size
companies with total assets ranging between 100 and 500 millions MYR. Twenty per
cent of the companies are large (between 500 and 1000 million MYR), and one quarter
of the sample companies are very large corporations (total assets .1000 millions
MYR). The oldest company was established in 1973, and several were started in 2005.
Table II categorizes the companies based on the foundation year. Companies with an
age of more than fteen years are nearly one third of the sample (95 firms 32 per
cent) whilst 46 per cent (137) of the companies in the sample were established during
the 1990s and in 2000. Sixty-eight rms (23 per cent) entered the market recently (after
year 2000). Table III illustrates the number of companies in each sector. The banks and
nancial rms represent 9 per cent of the sample companies while industrial
companies represent nearly half the sample. Table IV exhibits the minimum,
maximum, average and standard deviation of the DPS each year.
There are companies that did not pay dividends while the maximum dividend paid
to a share was 2.48 MYR in year 4. The means for the ve years (2001-2005) increased
from one year to the next (0.047-0.069) with a total increase of 46 per cent. The
Standard Deviation also increased by 11 per cent, which means that the dividend
paying variation among companies expanded. Table V represents the number and
percentage of companies categorized by the amount of MYR paid each share.
Size N %
,100 M 14 0.05
100 M-500 M 145 0.5
500M-1,000 M 56 0.20
.1,000 M 73 0.25
Total 288 100
Table I.
Sample companies
features
Sector N %
Banks and nancial sector 26 0.09
Real estate 43 0.14
Industrial sector 147 0.49
Others 84 0.28
Total 300 1.00
Table III.
Sample companies
features
Year of foundation N %
Before 1990 95 0.32
1990-2000 137 0.46
After 2000 68 0.23
Total 300 100
Table II.
Sample companies
features
JRF
8,4
352
The number of companies paying no dividends was 67 (30 per cent) in 2001. This
gure declined to 56 (24 per cent) by 2005. Seventy per cent of the sampled companies
paid dividends, 106 rms (68 per cent) of which 56 paid less than 0.05 MYR in 2001
with the number declining to 97 rms (54 per cent) by 2005. On the other hand, only 21
rms (13 per cent of paid dividend companies) paid more than 0.10 MYR per share in
2001 and this number doubled to 43 rms (24 per cent) in year 2005. This nding is in
contrast to the Fama and French (2001)s nding based on the USA market. Table VI
shows descriptive information about the net earnings per share during a ve-year
period.
The minimum earning (loss) was 211.82 in 2002 and declined to 22.61 in 2005. The
maximum, however, had not changed much during this period with the highest pay at
2.74 MYR in 2004. The mean of net EPS was small (0.015) in 2001 and even smaller
(0.005) in 2002. It then started improving during the period to reach 0.101 in 2005 to
reach a six-fold total increase. The SD suggests that the variance among companies
decreased in the last two years. Table VII represents the number and percentage of
companies categorized by the net EPS.
29 per cent of the sample companies were reporting negative Net EPS. However the
number of companies suffering high losses (20.10 MYR Net EPS or more) declined
from 38 (17 per cent) in 2001 to 30 (10 per cent) in 2005. The number of companies that
reported smaller amounts (less than 0.10 MYR) of negative Net EPS was 26 (12 per
N Minimum Maximum Mean SD
DPS 2001 224 0 1.954 0.047 0.141
DPS 2002 244 0 2.16 0.052 0.166
DPS 2003 261 0 2.398 0.053 0.159
DPS 2004 276 0 2.484 0.058 0.168
DPS 2005 237 0 2.068 0.069 0.156
Table IV.
Descriptive statistics of
dividends paid by sample
companies
2001 2002 2003 2004 2005
Dividends N % N % N % N % N %
DPS 0 67 30 76 31 82 31 84 30 56 24
0 , DPS, 0:05 106 47 116 48 113 43 110 40 97 41
0 , DPS, 0:10 30 13 26 11 32 12 47 17 41 17
DPS . 0.10 21 9 26 11 34 13 35 13 43 18
Total 224 244 261 276 237
Table V.
Number and percentage
of companies that paid or
did not pay dividends
. N Minimum Maximum Mean SD
Net EPS 2001 223 26.367 2.13 0.015 0.535
Net EPS 2002 244 211.824 2.322 0.005 0.881
Net EPS2003 261 26.507 2.655 0.060 0.598
Net EPS 2004 280 22.203 2.739 0.112 0.353
Net EPS 2005 287 22.611 2.076 0.101 0.320
Table VI.
Descriptive statistics of
net earnings per share
Dividend policy
and payout ratio
353
cent) in year 2001. That gure did not change much during the latter years. Moreover,
rms achieving low net prots (less than 0.10 MYR) yet also high prots per share
(0.10 MYR or more) rose from 72 (32 per cent), 88 (39 per cent), respectively, in 2001 to
102 (35 per cent), 132 (46 per cent), respectively, in 2005. Table VIII compares the
means of DPS for each corresponding pair of years.
As the t-test advocates, there are no signicant differences between the means of
DPS in years 2001 and 2002, 2002 and 2003, and 2004 and 2005. However, there is
signicant ( p , 0.05) difference between the means of DPS in years 2003 and 2004.
Statistical analysis
The literature suggests that a strong relationship exists between EPS and DPS. It also
suggests that the relationship between current DPS and previous or future DPS is high.
Table IX examines these relations via the Pearson Correlation.
Correlation coefcients show that the relationship between current DPS and past
DPS is signicantly (p 0:000) strong (Pearson . 0:80). Also, the relationship
between current EPS and previous EPS is signicant ( p , 0.05) in all cases except
between year 2005 and both 2001 and 2002. From the above coefcients, the correlation
between EPS in years 2001 and 2002 is strong (Pearson .0.90), but weaker in years
2001 and 2003, 2001 and 2004, 2002 and 2003, 2002 and 2004, and even weaker between
2003 and 2004, and 2004 and 2005. The weakest signicant relation exists between
2003 and 2005 (Pearson ,0.16). The correlations between DPS and EPS for all 5 years
is signicant ( p , 0.01) with strong (Pearson .0.50) relations existing between EPS
for 2004 and DPS for all years, and between DPS and EPS in the fth year. From these
results, we can conclude that companies, in general, follow stable dividend policies not
strongly attached to current earnings. These ndings are consistent with Baker et al.
(2001). Based on these correlation signicances, we reject both H1 and H2 while we
accept the alternative. That is, a strong relationship exists between: current DPS, its
past and future; and current DPS, current EPS and its past and future.
Pair years Mean SD t df Sig. (2-tailed)
2001-2002 0.007 0.076 21.404 222 0.162
2002-2003 20.003 0.081 20.647 242 0.518
2003-2004 20.007 0.056 21.991 257 0.047
2004-2005 20.004 0.069 20.881 225 0.379
Note: Sig. means p-value
Table VIII.
DPS yearly comparison
2001 2002 2003 2004 2005
N % N % N % N % N %
Net EPS , 20:10 38 17 41 17 27 10 32 11 30 10
0 , Net EPS ,20.10 26 0.12 20 0.08 21 0.08 20 0.07 24 0.08
0.10 . Net EPS .0 72 32 74 30 86 33 82 29 102 35
Net EPS. 0:10 88 39 110 45 128 49 147 52 132 46
Total 224 245 262 281 288
Table VII.
Number and percentage
of companies categorized
by net earnings per share
JRF
8,4
354
D
P
S
D
P
S
D
P
S
D
P
S
D
P
S
N
e
t
E
P
S
N
e
t
E
P
S
N
e
t
E
P
S
N
e
t
E
P
S
N
e
t
E
P
S
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
2
0
0
5
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
2
0
0
5
D
P
S
2
0
0
1
P
e
a
r
s
o
n
c
o
r
r
e
l
a
t
i
o
n
1
0
.
9
0
3
0
.
9
6
9
0
.
9
4
5
0
.
9
2
0
0
.
3
4
7
0
.
2
3
4
0
.
4
2
9
0
.
5
3
2
0
.
4
5
7
S
i
g
.
(
2
-
t
a
i
l
e
d
)
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
N
2
2
3
2
2
3
2
2
2
1
8
9
2
2
3
2
2
3
2
2
3
2
2
3
2
2
3
D
P
S
2
0
0
2
P
e
a
r
s
o
n
c
o
r
r
e
l
a
t
i
o
n
1
0
.
8
8
0
0
.
9
5
1
0
.
8
3
0
0
.
3
4
4
0
.
2
3
0
0
.
4
0
5
0
.
5
0
5
0
.
4
2
4
S
i
g
.
(
2
-
t
a
i
l
e
d
)
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
N
2
4
3
2
4
2
2
0
4
2
2
3
2
4
3
2
4
3
2
4
3
2
4
2
D
P
S
2
0
0
3
P
e
a
r
s
o
n
c
o
r
r
e
l
a
t
i
o
n
1
0
.
9
4
7
0
.
9
2
8
0
.
3
4
7
0
.
2
3
0
0
.
3
4
1
0
.
5
4
5
0
.
4
6
7
S
i
g
.
(
2
-
t
a
i
l
e
d
)
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
N
2
5
8
2
1
3
2
2
3
2
4
4
2
6
0
2
6
0
2
5
7
D
P
S
2
0
0
4
P
e
a
r
s
o
n
c
o
r
r
e
l
a
t
i
o
n
1
0
.
9
2
8
0
.
3
5
2
0
.
2
3
6
0
.
3
3
9
0
.
5
6
6
0
.
4
8
5
S
i
g
.
(
2
-
t
a
i
l
e
d
)
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
N
2
2
6
2
2
2
2
4
3
2
5
9
2
7
5
2
7
1
D
P
S
2
0
0
5
P
e
a
r
s
o
n
c
o
r
r
e
l
a
t
i
o
n
1
0
.
3
5
0
0
.
2
4
1
0
.
3
5
1
0
.
5
7
2
0
.
5
2
9
S
i
g
.
(
2
-
t
a
i
l
e
d
)
0
.
0
0
0
0
.
0
0
1
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
N
1
8
9
2
0
5
2
1
4
2
2
7
2
3
7
N
e
t
E
P
S
2
0
0
1
P
e
a
r
s
o
n
c
o
r
r
e
l
a
t
i
o
n
1
0
.
9
0
2
0
.
5
2
5
0
.
5
6
6
0
.
0
7
3
S
i
g
.
(
2
-
t
a
i
l
e
d
)
0
.
0
0
0
0
.
0
0
0
0
.
0
0
0
0
.
2
7
6
N
2
2
3
2
2
3
2
2
3
2
2
3
N
e
t
E
P
S
2
0
0
2
P
e
a
r
s
o
n
c
o
r
r
e
l
a
t
i
o
n
1
0
.
5
4
6
0
.
5
3
9
0
.
0
4
0
S
i
g
.
(
2
-
t
a
i
l
e
d
)
0
.
0
0
0
0
.
0
0
0
0
.
5
4
0
N
2
4
4
2
4
4
2
4
3
N
e
t
E
P
S
2
0
0
3
P
e
a
r
s
o
n
c
o
r
r
e
l
a
t
i
o
n
1
0
.
4
4
5
0
.
1
5
9
S
i
g
.
(
2
-
t
a
i
l
e
d
)
0
.
0
0
0
0
.
0
1
1
N
2
6
1
2
5
8
N
e
t
E
P
S
2
0
0
4
P
e
a
r
s
o
n
c
o
r
r
e
l
a
t
i
o
n
1
0
.
3
3
4
S
i
g
.
(
2
-
t
a
i
l
e
d
)
0
.
0
0
0
N
2
7
4
N
e
t
E
P
S
2
0
0
5
P
e
a
r
s
o
n
c
o
r
r
e
l
a
t
i
o
n
1
S
i
g
.
(
2
-
t
a
i
l
e
d
)
N
2
8
7
Table IX.
Pearson correlation
results on DPSs and EPS
Dividend policy
and payout ratio
355
Effect of company features
What other factors that could have some effects on dividends? Table X compares the
DPS for three years (2001, 2003 and 2005) between companies established before 1990
and companies founded later.
Old companies paid better dividends per share than newer companies. However, the
differences were not signicant at the 10 per cent level. In years 2001 and 2005, the
difference is signicant at 15 per cent. Examining for sector (industry vs. other sectors)
impact shows no signicant ( p . 0.10) difference in the means of DPS for all ve
years. Thus we do not reject the null hypothesis that companys age or sector has no
effect on paid DPS. Table XI compares the DPS between large (Total assets . 1; 000
Millions MYR) and small (,1000 Millions MYR).
The large companies gave signicantly ( p , 0.10) better DPS in 2001. In 2003,
however, the difference between DPS in large and small companies increased and was
signicant at the 5 per cent level. This difference also increased even more (0.073 MYR)
in the fth year and was more signicant ( p , 0.01). Therefore, we do not accept the
hypothesis that the company size has no effect on the DPS.
Payout ratio
The payout ratio reecting the percentage of net income (available for shareholders)
can follow different policies. Some companies prefer xed (or semi-xed) payout ratios
while some others choose to pay xed amounts (with a small annual increase)
regardless of the gained net income. Table XII shows the number of companies (and
percentages) in each of four categories.
The negative payout ratio means that companies pay dividends even though they
report losses. In 2001, 8 per cent of the sample companies gave out dividends without
earning prots and even while reporting losses. However this percentage decreased to
4 per cent in year 2005. This decrease probably reects nancial difculties facing
these companies. Twenty-nine per cent of the rms were paying out less than one third
of their prot and this percentage declined in year 2005 to only 16 per cent. Not many
Year Total assets (000) N Mean SD t df Sig. (2-tailed)
2001 . 1; 000; 000 66 0.072 0.242 1.698 221 0.091
,1,000,000 157 0.037 0.059
2003 . 1; 000; 000 68 0.091 0.294 2.311 258 0.022
,1,000,000 192 0.040 0.059
2005 . 1; 000; 000 60 0.124 0.273 3.216 235 0.001
,1,000,000 177 0.051 0.077
Table XI.
DPS means comparison
between small and large
rms
Date est. N Mean SD t df Sig. (2-tailed)
DPS2001 . 1990 143 0.037 0.048 21.580 219 0.116
,1990 78 0.068 0.228
DPS2003 . 1990 170 0.045 0.063 21.118 256 0.265
, 1990 88 0.069 0.260
DPS2005 . 1990 160 0.059 0.079 21.444 234 0.150
,1990 76 0.091 0.249
Table X.
DPS means comparison
using a 1990 year cutoff
JRF
8,4
356
companies, only 18 per cent in 2001 and 21 per cent in 2005, paid out half or more of
their prots. About a third of the companies were paid no dividends. Table XIII
correlates payout ratios to net prots.
Generally, there is no signicant correlation existing between earnings and payout
ratios. The only signicant relationship exists between the net prot of 2004 and the
2003 payout ratio and the net prot of 2001 and the 2002 payout ratio. This conrms
the previous results that most Malaysian companies have dividend payout policies that
are not obviously connected to the companys income for the year. An examination for
correlations between the payout ratio and sales per share shows no signicant
relationship ( p . 0.30) except for 2003 (p 0:07). Also a test for correlation between
the payout ratio and the published cash-per-share revealed no signicant ( p . 0.65)
correlations except for 2003 as well ( p , 0.05). Again we looked at the correlations
between the payout ratio and the book-value per share. There were no signicant
(p . 0:50) relationships. In addition, we were unable to determine any signicant
correlation between the payout ratio and the future return on equity ( p . 0.45). From
these results we cannot reject H5. Thus we accept that POR does not, in general terms,
have a signicant impact on the companys future growth. Company size has no
signicant ( p . 0.75) impact on the payout ratio. There are also no signicant
( p . 0.20) differences in the means of POR between industrial (136 rms) and
non-industrial (135 rma) sectors. A Pearson correlation shows a negative relationship
Payout ratio
2001 2002 2003 2004 2005
Net EPS 2001 Pearson correlation 0.048 0.134 0.097 0.038 0.002
Sig. (2-tailed) 0.492 0.057 0.157 0.582 0.975
N 203 203 213 213 212
Net EPS 2002 Pearson correlation 0.047 0.088 0.072 0.033 20.001
Sig. (2-tailed) 0.507 0.190 0.276 0.614 0.989
N 203 224 233 234 232
Net EPS 2003 Pearson correlation 0.045 0.079 0.082 0.034 20.004
Sig. (2-tailed) 0.525 0.240 0.198 0.594 0.946
N 203 224 250 251 246
Net EPS 2004 Pearson correlation 0.069 0.099 0.132 0.025 20.010
Sig. (2-tailed) 0.330 0.140 0.036 0.687 0.875
N 204 225 252 268 261
Net EPS 2005 Pearson correlation 0.027 0.057 0.088 0.024 20.021
Sig. (2-tailed) 0.704 0.396 0.165 0.694 0.726
N 203 224 249 263 274
Table XIII.
Correlations between
payout ratios and net
prot per share
Payout ratio (POR) 2001 2002 2003 2004 2005
N % N % N % N % N %
POR , 0 20 8 21 7 11 4 10 3 11 4
0 , POR , 0.33 72 29 89 32 84 28 76 25 47 16
0.33, POR , 0.50 25 10 26 9 49 17 52 17 32 11
. 0:50 44 18 53 19 51 17 44 15 62 21
Total 247 281 296 300 300
Table XII.
Number and percentage
of companies categorized
by payout ratio
Dividend policy
and payout ratio
357
between the payout ratio and the borrowing ratio, but these relationships are not
signicant ( p . 0.10). The working capital ratio was not found to be signicantly
correlated to the payout ratio ( p . 0.10) except for 2004 (sig: 0:03). Table XIV
represents the correlation matrix between current and past payout ratios.
The correlation is signicant ( p , 0.05) between 2001 and each of the subsequent
three years while the correlation is not signicant with 2005 ( p . 0.75). Similarly, the
correlation between 2002 and each of the subsequent three years is signicant
(p 0:000). Also, 2003 has a signicant link ( p , 0.10) with both 2004 and 2005.
Nevertheless, the correlation between the fourth year 4 and the fth year is not
signicant. Based on these results, we can reject H6 and agree that POR is strongly
affected by its past and affects its future. By comparing the means of POR between
each two sequencing years, we found that the payout ratio does not signicantly
( p . 0.30) differ from one year to the following except between years 2001 and 2002
since there is signicant change in the payout ratio (p 0:05). Table XV reveals the
relationship between the payout ratio and the capital-gearing ratio.
The above coefcient signs suggest that companies with high leverage pay lower
dividends from their prots and vice versa. The 2001 capital-gearing ratio has a
negatively signicant effect on the payout ratios of the same year and the following
three years. As though we may not be able to accept Null H7, namely that the payout
ratio is not negatively related to the companys leverage level. Table XVI compares
dividend payout ratios between companies achieving 0.05 MYR EPS or more and other
companies with EPS less than 0.05 MYR.
As the POR means suggest, in 2001, companies with low EPS tend to distribute
most earnings (m 0:74), but companies with high EPS distribute an average of only
one third of earnings. The difference between the means of these two groups of
companies is signicant at the 5 per cent level. However, this strategy has changed
over years since in 2005 the difference of the POR means between high and low earning
companies was minor and insignicant (p 0:65). Another possible factor having an
impact upon the size of POR is the cash available per share. Table XVII shows the
statistical results of the mean comparison of the DPS and POR of two groups of
companies (large and small amounts of available cash).
Payout ratio
Payout ratio 2002 2003 2004 2005
2001 Pearson correlation 0.248 0.141 0.204 0.021
Sig. (2-tailed) 0.001 0.049 0.004 0.773
N 191 195 194 195
2002 Pearson correlation 0.373 0.329 0.310
Sig. (2-tailed) 0.000 0.000 0.000
N 218 217 213
2003 Pearson correlation 0.118 0.118
Sig. (2-tailed) 0.065 0.069
N 247 240
2004 Pearson correlation 0.045
Sig. (2-tailed) 0.478
N 255
Table XIV.
Correlations between
current and past payout
ratios
JRF
8,4
358
Companies which had a larger amount of cash available for shares gave higher
dividends (m 0:077 per share in 2001 and 0.102 in 2005) while the other companies
which had less cash available tended to pay at a much lower rate (m 0:014 per share
in 2001 and 0.018 in 2005). The differences between these means are signicant at the 1
per cent level. The POR was surprisingly higher in those companies with lower cash
available in 2001 but the difference (0.27) is not signicant using 10 per cent level. The
situation had changed reversely in 2005 since the POR in companies with larger
available cash (m 0:413) is signicantly ( p , 0.01) higher than it in the other
companies (m 0:245). Table XVIII provides comparison analysis of DPS and POR
based on the share book value.
Payout ratio Capital-gearing ratio
2001 2002 2003 2004 2005
2001 Pearson correlation 20.119 20.064 20.006 20.014 20.127
Sig. (2-tailed) 0.090 0.365 0.928 0.845 0.071
N 203 202 201 203 203
2002 Pearson correlation 20.214 20.102 20.024 20.051 20.221
Sig. (2-tailed) 0.001 0.127 0.723 0.444 0.001
N 218 224 222 224 223
2003 Pearson correlation 20.160 20.085 20.017 20.036 20.144
Sig. (2-tailed) 0.015 0.192 0.787 0.574 0.023
N 228 240 247 251 248
2004 Pearson correlation 20.115 20.042 20.009 20.043 20.131
Sig. (2-tailed) 0.083 0.519 0.881 0.484 0.034
N 227 243 257 268 263
2005 Pearson correlation 0.032 0.000 0.007 0.019 0.039
Sig. (2-tailed) 0.634 0.999 0.918 0.759 0.524
N 225 241 253 271 274
Table XV.
Pearson correlation
between the payout ratio
and the capital-gearing
ratio
Year Net EPS N Mean SD t df Sig. (2-tailed)
2001 . 0:050 122 0.329 0.286 22.401 201 0.017
,0050 81 0.738 1.853
2005 . 0:050 178 0.352 0.359 0.459 271 0.646
,0.050 95 0.323 0.695
Table XVI.
Mean comparison of
payout ratios between
rms using a net EPS
0.05 cutoff
Year Variable Cash per share N Mean SD t df Sig. (2-tailed)
2001 DPS . 0:100 120 0.077 0.186 3.350 219 0.001
,0.100 101 0.014 0.022
POR . 0:100 111 0.377 0.437 21.565 199 0.119
,0.100 90 0.645 1.734
2005 DPS . 0:100 144 0.102 0.190 4.202 235 0.000
,0.100 93 0.018 0.041
POR . 0 0:100 157 0.413 0.501 2.782 271 0.006
,0.100 117 0.245 0.486
Table XVII.
Mean comparison of DPS
and payout ratios of rms
using a net-cash-per-share
0.1 cutoff
Dividend policy
and payout ratio
359
The shares that have higher book values got more dividends (average 0.07-0.102)
than shares with lower book values (average 0.025-0.035). The differences between
these means (in years 2001 and 2005) are signicant ( p , 0.05). Payout ratios were also
higher for shares with higher book values. The difference in the means was not
signicant in 2001 but became signicant in 2005. The standard deviation had fallen
from 1.45 and 0.92 to 0.46 and 0.53, and this meant that the variation of POR between
companies heavily decreased.
The payout ratio and the companys future earning growth
Until recently, it was generally accepted that companies that reserve more of their
earnings for future growth are expected to increase their earnings in future and vice
versa. Recently, however, researchers have challenged this theory. They found a
positive signicant relation connection between future growth in earnings and the
payout ratio. These studies were based on companies from developed industries. In
this part, we are examining this issue using companies from Malaysia, which is a
semi-developed nation (Thanoon et al., 2005). Examining the relationship between the
payout ratio and future earnings growth could either support the old presumption, the
recent ndings or neither. Using the changes of EPS as a proxy of one to four years of
future earnings growth, the results are represented in Table XIX.
As indicated in the table above, the relationship between POR and future earnings
growth is negative but insignicant. This means that the POR had no real impact on
Year Variable Share book value N Mean SD t df Sig. (2-tailed)
2001 DPS . 1:500 112 0.070 0.188 2.390 221 0.018
,1.500 111 0.025 0.057
POR . 1:500 97 0.618 1.449 1.434 201 0.153
,1.500 106 0.376 0.916
2005 DPS . 1:500 121 0.102 0.200 3.427 235 0.001
,1.500 116 0.035 0.076
POR . 1:500 128 0.422 0.456 2.509 271 0.013
,1.500 145 0.271 0.529
Table XVIII.
Mean comparison of DPS
and payout ratios of rms
using a net share
book-value 1.5 cutoff
Future earning growth (change in EPS)
POR 1 year 2 years 3 years 4 years
2001 Pearson correlation 0.035 20.016 20.002 20.027
Sig. (2-tailed) 0.616 0.826 0.978 0.704
N 203 203 203 203
2002 Pearson correlation 20.064 20.056 20.063
Sig. (2-tailed) 0.343 0.406 0.351
N 224 224 223
2003 Pearson correlation 20.002 20.031
Sig. (2-tailed) 0.977 0.623
N 250 247
2004 Pearson correlation 20.002
Sig. (2-tailed) 0.977
N 263
Table XIX.
Pearson correlation of the
payout ratio and future
earnings growth as
measured by EPS
JRF
8,4
360
future earnings. This contrasts both with the traditional wisdom and with the recent
ndings of empirical research based on developed markets (Arnott and Asness, 2003;
Zhou and Ruland, 2006; Gwilym et al., 2006). To verify this result, we correlated the
POR to the future growth on sales per share, and we reached the same conclusion. Also
by running the correlation analysis between the POR and the future changes on return
on equity (ROE), we get similar results. Controlling for the companys age, size and
industry did not change the above results signicantly. Using a different analysis
strategy, OLS regression estimation leads to the same conclusion. Therefore, we cannot
reject H8 , that the payout ratio has no signicant impact on company future earnings
growth.
These ndings might help inexperienced managers in the emerging markets to
understand the dividend policy followed by well-established corporations. As reported,
the companies were following a stable, or semi-stable EPS strategy (not strictly tied to
earnings), and this could be better implemented within the Malaysian region or
probably in all of South-East Asia. Also, since there is no evidence that the POR was
signicantly correlated to the companys future growth, retaining the earnings, or most
of them, might be a conventional alternative.
Conclusions
A dividend policy and a payout ratio could make signicant impact on the corporate
future value when well established and carefully followed. Corporate governance
should institute an effective mechanism of how much to pay as share dividends and
when to pay, taking into account a variety of factors relating to the companys current
status, its future as well as market and economic circumstances. This paper endeavors
to highlight the factors that might have an effect on the dividend policy and/or the
payout ratio. The results suggest that around 70 per cent of the surveyed companied
have paid dividend during the last ve years (2001-2005), whilst the mean of EPS
increased from 0.015 in year 2001 to over 0.10 in year 2005 (6 times). The means
comparison from year to the following shows no signicant differences of EPS except
between 2003 and 2004.
Statistical analysis was performed to test the eight hypotheses. Pearson correlation
results conrmed what was suggested in the literature that current dividends are
affected by its past and future. Also dividends were associated with net earnings but
less strongly. Neither the age of the paying dividend company nor its home sector
(industry and non-industry) has an impact on the amount paid for each share (DPS).
However, size has a signicant effect on the DPS. In contrast to previous studies, the
payout ratios (POR) have no signicant association with either the current, past or
future net earnings, and the correlation between POR and the companys future
earning growth, in general, is negative but insignicant. On the other hand, the
companys level of leverage has a negative (sometimes signicantly) relationship with
the POR. The strongest determiner of the POR is its past ratios. The year 2001 mean
comparison of payout ratios between companies with net EPS 0.05 MR or more and the
other companies showed surprisingly that companies with less EPS distribute most
earnings (average 74 per cent) as dividends while the other companies with higher
earnings per share paid only one third of the earnings to shareholders. That difference
is signicant. Also, a means comparison analysis revealed that cash per share has a
signicant positive impact not only on DPS but also on POR. In addition to that, a t-test
Dividend policy
and payout ratio
361
of mean comparisons suggests that shares with higher book values receive
signicantly more dividends, and that the POR is signicantly higher for those shares.
It is probably the time for comparison studies between markets with different levels
of development since different markets require different dividend policies. Future
studies might also examine to what extent dividend strategy can be affected by culture,
economic status, and commercial environment. On the other hand, investors views
regarding their preference on dividends should be surveyed, and this might assist
company corporate governance in establishing good dividend plans.
References
Amidu, M. and Abor, J. (2006), Determinants of dividend payout ratios in Ghana, Journal of
Risk Finance, Vol. 7 No. 2, pp. 136-45.
Arnott, R. and Asness, C. (2003), Surprise! Higher dividends higher earnings growth,
Financial Analysts Journal, Vol. 59 No. 1, pp. 70-87.
Baker, H. (1988), Relationship between industry classication and dividend policy, Southern
Business Review, Vol. 14 No. 1, pp. 1-8.
Baker, H. and Powell, G. (2000), Determinants of corporate dividend policy: a survey of NYSE
rms, Financial Practices & Education, Spring/Summer, pp. 29-40.
Baker, H., Farrelly, G. and Edelman, R. (1985), A survey of management views on dividend
policy, Financial Management, Vol. 14 No. 3, pp. 78-84.
Baker, H., Veit, E. and Powell, G. (2001), Factors inuencing dividend policy decisions of
Nasdaq rms, The Financial Review, Vol. 36 No. 3, pp. 19-37.
Benartzi, S., Michaely, R. and Thaler, R. (1997), Do changes in dividends signal the future or the
past?, Journal of Finance, Vol. 52 No. 3, pp. 1007-34.
DeAngelo, H., DeAngelo, L. and Skinner, D. (2004), Are dividends disappearing? Dividend
concentration and the consolidation of earnings, Journal of Financial Economics, Vol. 72,
pp. 425-56.
Dickens, R., Casey, K. and Newman, J. (2002), Bank dividend policy: explanatory factors,
Quarterly Journal of Business & Economics, Vol. 41, pp. 3-12.
Fama, E. and French, K. (2001), Disappearing dividends: changing rm characteristics or lower
propensity to pay?, Journal of Financial Economics, Vol. 60 No. 2, pp. 3-43.
Fama, F. (1974), The empirical relationship between the dividend and investment decisions of
rms, American Economic Review, Vol. 64 No. 3, pp. 304-18.
Gordon, M. (1959), Dividends, earnings and stock prices, Review of Economics and Statistics,
Vol. 41, pp. 99-105.
Gwilym, O., Seaton, J., Suddason, K. and Thomas, S. (2006), International evidence on the payout
ratio, earnings, dividends, and returns, Financial Analysts Journal, Vol. 62 No. 1, pp. 36-53.
Lintner, J. (1956), Distribution of incomes of corporations among dividends, retained earnings,
and taxes, American Economic Review, Vol. 61, pp. 97-113.
Mancinelli, M. and Ozkan, A. (2006), Ownership structure and dividend policy: evidence from
Italian rms, European Journal of Finance, Vol. 12 No. 3, pp. 265-382.
Miller, M. and Modigliani, F. (1961), Dividend policy, growth, and the valuation of shares,
Journal of Business, Vol. 34 No. 4, pp. 411-33.
Pruitt, S. and Gitman, L. (1991), The interactions between the investment, nancing, and
dividend decisions of major US rms, Financial Review, Vol. 26 No. 3, pp. 409-30.
JRF
8,4
362
Singhania, M. (2005), Trends in dividend payout: a study of select Indian companies, Journal of
Management Research, Vol. 5 No. 3, pp. 129-42.
Thanoon, M., Baharumshah, A. and Rahman, A. (2005), Malaysia: from economic recovery to
sustained economic growth, Journal of Post Keynesian Economics, Vol. 28 No. 2,
pp. 295-315.
Zhou, P. and Ruland, W. (2006), Dividend payout and future earnings growth, Financial
Analysts Journal, Vol. 62 No. 3, pp. 58-69.
Corresponding author
Abdulrahman Ali Al-Twaijry can be contacted at: atwaijry@lycos.com
Dividend policy
and payout ratio
363
To purchase reprints of this article please e-mail: reprints@emeraldinsight.com
Or visit our web site for further details: www.emeraldinsight.com/reprints
Debt policy and performance
of SMEs
Evidence from Ghanaian and South African
rms
Joshua Abor
University of Stellenbosch Business School, South Africa, and
University of Ghana Business School, Legon, Ghana
Abstract
Purpose The purpose of this research is to examine the effect of debt policy (capital structure) on
the nancial performance of small and medium-sized enterprises (SMEs) in Ghana and South Africa.
Previous studies, especially on large rms, have shown that capital structure affects rm performance.
Though the issue has been widely studied, largely missing from this body of literature is the focus on
SMEs.
Design/methodology/approach Panel data analysis is used to investigate the relations between
measures of capital structure and nancial performance.
Findings Using various measures of performance, the results of this study indicate that capital
structure inuences nancial performance, although not exclusively. By and large, the results indicate
that capital structure, especially long-term and total debt ratios, negatively affect performance of
SMEs. This suggests that agency issues may lead to SMEs pursuing very high debt policy, thus
resulting in lower performance.
Originality/value The main value of this paper is the analysis of the effect of debt policy on the
performance of SMEs in Ghana and South Africa.
Keywords Debts, Capital structure, Small to medium-sized enterprises, Ghana, South Africa
Paper type Research paper
1. Introduction
One important nancial decision rms are confronted with is the debt policy or capital
structure choice. This decision is particularly crucial given the effect it has on the value
of the rm. The capital structure of a rm is a specic mix of debt and equity the rm
uses to nance its operations (Abor, 2005). In general, a rm can choose among many
alternative capital structures. It can issue a large amount of debt or very little debt. It
can arrange lease nancing, use warrants, issue convertible bonds, sign forward
contracts or trade bond swaps. It can issue dozens of distinct securities in countless
combinations. It is important for the rm however, to nd the particular combination
of debt and equity that maximizes its overall market value. Managers who are astute
enough to identify and deploy the appropriate mix of debt and equity are amply
rewarded in the market place, because, all things being equal, this appropriate mix of
debt and equity minimizes a rms cost of nancing. Given revenue and prenancing
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1526-5943.htm
An earlier version of this paper was presented at the Seventh International Academy of African
Business and Development Conference at the Ghana Institute for Management and Public
Administration, Ghana, May, 2006.
JRF
8,4
364
The Journal of Risk Finance
Vol. 8 No. 4, 2007
pp. 364-379
qEmerald Group Publishing Limited
1526-5943
DOI 10.1108/15265940710777315
prot streams that are generated through non-nancial factors, minimizing cost of
nancing maximizes net returns for the rm, thereby improving its competitive
advantage in the marketplace (Gleason et al., 2000). It is suggested that utilization of
different levels of debt and equity in the rms capital structure is one such
rm-specic strategy used by managers in search for improved performance (Gleason
et al., 2000).
This interplay of debt and equity and corporate performance has been the subject of
a number of studies. Such empirical studies on the effect of capital structure on
protability have tended to concentrate on large rms (see Krishnan and Moyer, 1997;
Majumdar and Chhibber, 1999; Abor, 2005). Previous empirical studies on SMEs,
though limited, have also focused on the determinants of capital structure. Amajor gap
in the literature is the examination of the effect of capital structure on the performance
of SMEs. This present study examines the effect of debt policy on nancial
performance of SMEs. Using data of Ghanaian SMEs and South African SMEs for a
period of six years, panel regression model is employed for this study. The study also
limited the sample to quoted South African SMEs in order to evaluate the effect of the
debt policy using the Tobins q as a measure of performance. The results of the study
generally suggest that agency conicts may be largely responsible for the excessive
use of debt by SMEs, leading to negative relationship between capital structure and
nancial performance.
The rest of the paper is organized as follows: The next session gives a review of the
extant literature on the subject. Section three describes the methodology used for this
study. Section four presents and discusses the results of the empirical analysis. Finally,
section ve summarizes the ndings of the research and also concludes the discussion.
2. Literature review
Recent theory on capital structure is based on the Modigliani and Millers (1958)
seminal work on the effect of capital structure on the value of the rm. Their theory
assumes perfect markets and perfect competition in which rms operate without taxes
or transaction cost and where all relevant information is available without cost.
However, these assumptions do not hold in the real world or in practice and factors
such as taxes, agency cost, cost of nancial distress and information asymmetry are
important in explaining the capital structure of rms.
Modigliani and Miller have been criticized on the grounds that their theory assumes
rational economic behaviour and perfect markets conditions, owners goals are
targeted only at maximizing prots (Grabowksi and Mueller, 1972), and that it has
limited applicability to small rms (Chaganti et al., 1995). Modigliani and Miller (1963)
revised their former stance by incorporating tax benets as determinant of the rms
capital structure choice. They argue that rms are able to maximize their value by
employing more debt because of the tax-shield benets associated with debt use.
Interest on debt is considered as a tax-allowable expense. Some researchers have
subsequently suggested alternatives to the Modigliani and Miller theory of capital
structure by including the agency theory (Jensen and Meckling, 1976), the bankruptcy
cost (Titman, 1984) and the pecking order theory (Myers, 1984; Myers and Majluf,
1984). The extant literature offers different perspectives about how the decision to
acquire debt affects rm value.
Debt policy
365
Hutchinson (1995) argues that in more general terms, nancial leverage has a
positive effect on the rms return on equity provided that earnings power of the rms
assets (the ratio of earnings before interest and taxes to total assets) exceeds the
average interest cost of debt to the rm. He argues that the extent to which a rms
earnings power is likely to remain above the breakeven point and the potential speed
or exibility with which it can adjust its debt usage, if its earnings power falls below
average interest costs, should help to determine the level of debt that the rm is willing
to commit itself to at a given point in time. Taub (1975) found signicantly positive
relationship between debt ratio and measures of protability. Nerlove (1968), Baker
(1973), and Petersen and Rajan (1994) also identied positive association between debt
and protability but for industries. In their study of leveraged buyouts, Roden and
Lewellen (1995) established a signicantly positive relation between protability and
total debt as a percentage of the total buyout-nancing package. Champion (1999)
pointed out that the use of leverage was one way to improve the performance of the
rm. Hadlock and James (2002) also concluded that companies prefer debt nancing
because they anticipate higher returns. It is believed that large debt holders have an
interest in seeing that managers take performance-improving measures. Kaplan and
Minton (1994), and Kang and Shivdasani (1995) found higher incidence of management
turnover in Japan in response to poor performance in companies that have a principal
banking relationship relative to companies that do not.
Other studies such as those by Ross (1977), Heinkel (1982) and Noe (1988) suggest
that increasing leverage, by acquiring debt should, have positive implications for rm
value and performance. In general, these theories ascribe a signaling or disciplinary
role for debt. Since increasing debt would also increase bankruptcy and liquidation
costs, only managers who expect better future performance will choose to issue debt.
Graham and Harvey (2001) surveyed CFOs and report that managers are concerned
with maintaining nancial exibility and their rms credit rating when considering
debt issues. Since rm performance is frequently used as an input into the credit rating
decisions, this provides indirect survey evidence that managers issue debt keeping in
view of expected future performance. The agency model of Jensen (1986) suggests that
since debt sales bring additional cash into the rm, this could exacerbate agency
problems. Alternatively, if rms use the debt issue proceeds to address the gap
between investments needs and internal sources of funding, this would not necessarily
lead to an increase in excess cash within the rm. The periodic interest payments on
debt would then commit managers to pay out excess free cash ow. Hence, debt issues
could reduce agency costs, and have positive effects on rm value. In contrast, Miller
and Rock (1985), and Smith (1986) argue that all securities sales (including debt)
indicate decreases in future operating performance, and hence impact negatively on
rm value.
However, some studies have shown that debt has a negative effect on rm
protability. Fama and French (1998), for instance argue that the use of excessive debt
creates agency problems among shareholders and creditors and that could result in
negative relationship between leverage and protability. Majumdar and Chhibber
(1999) found in their Indian study that leverage has a negative effect on performance,
while Krishnan and Moyer (1997) connect capital and performance to the country of
origin. Gleason et al. (2000) support a negative impact of leverage on the protability of
the rm. In a polish study, Hammes (1998) also found a negative relationship between
JRF
8,4
366
debt and rms protability. In another study, Hammes (2003) examined the relation
between capital structure and performance by comparing Polish and Hungarian rms
to a large sample of rms in industrialized countries. He used panel data analysis to
investigate the relation between total debt and performance as well as between
different sources of debt namely, bank loans, and trade credits and rms performance
measured by protability. His results show a signicant and negative effect for most
countries. He found that the type of debt, bank loans or trade credit is not of major
importance, what matters is debt in general. Mesquita and Lara (2003), in their study
found that the relationship between rates of return and debt indicates a negative
relationship for long-term nancing. They however, found a positive relationship for
short-term nancing and equity. In a recent study, Abor (2005) examined the effect of
capital structure on the corporate protability of listed rms in Ghana using a panel
regression model. His measures of capital structure included short-term debt ratio,
long-term debt ratio and total debt ratio. His ndings show a signicantly positive
relation between the short-term debt ratio and protability. However, a negative
relationship between long-term debt ratio and protability was established. In terms of
the relationship between total debt ratio and protability, the results of his study
indicated a signicantly positive association between total debt ratio and protability.
In summary, empirical studies have given inconclusive results regarding the capital
structure choice and its effect on rms performance. This present study contributes to
the issue by investigating the effect of debt policy or capital structure on rm
performance by focusing on SMEs, which are often neglected in most empirical studies.
3. Research methodology
3.1 Data and measurement
This study sampled both Ghanaian and South African SMEs. The Ghanaian sample
was obtained from the databases of the National Board of Small Scale Industries and
the Association of Ghana Industries. The South African SMEs were sampled from the
register of the Small Business Advisory Bureau database. An SME in Ghana is dened
as a rm having less than 100 employees. This is based on the denition given by the
Regional Project on Enterprise Development for SMEs in Ghana. South African SMEs
are dened as rms that satisfy, at least, two of the following criteria; have less than
200 employees; turnover of less than 50 million South African Rand; gross assets
excluding xed property of less than 18 million South African Rand. This denition is
also consistent with that of the National Small Business Act for SMEs in South Africa.
The nancial data was obtained from the nancial statements of the rms for the six
year period, 1998-2003. In all 160 Ghanaian SMEs and 200 South African SMEs were
used for this study with 68 of the South African sample being listed rms.
The dependent variable is performance and the independent variables are the debt
ratios. Measures of nancial performance include, gross prot margin, return on assets
and for the listed SMEs, Tobins q. The debt ratios include short-term debt ratio,
long-term debt ratio and total debt ratio. Trade credit is also included to examine its
effects on performance. Trade credit is expected to have a positive impact on
performance. Trade creditors extend credit to rms with risky but positive net present
value (NPV) projects due to their superior knowledge, and higher ability to salvage
value as compared to other providers of debt nance and their ability to discipline
debtors by withholding future deliveries (Hammes, 2003). Two control variables (size
Debt policy
367
and growth) are also included as standard determinants of performance. The model for
the empirical investigation can be stated as follows:
Performance
i;t
b
0
b
1
SDC
i;t
b
2
FS
i;t
b
3
SG
i;t
m
it
1
Performance
i;t
b
0
b
1
LDC
i;t
b
2
FS
i;t
b
3
SG
i;t
m
it
2
Performance
i;t
b
0
b
1
TDC
i;t
b
2
FS
i;t
b
3
SG
i;t
m
it
3
Performance
i;t
b
0
b
1
TCC
i;t
b
2
FS
i;t
b
3
SG
i;t
m
it
4
where:
SDC
i,t
short-term debt/total capital for rm i in time t;
LDC
i,t
long-term debt/total capital for rm i in time t;
TDC
i,t
total debt/total capital for rm i in time t;
TCC
i,t
trade credit/total capital for rm i in time t
FS
i,t
rm size (log of total assets) for rm i in time t;
SG
i,t
log of sales growth for rm i in time t; and
m
i,t
the error term.
The performance measures are dened as; gross prot margin gross prot divided
by sales; return on assets net prot divided by total assets. Besides analyzing the
effect of the debt policy on the protability, the study also limited the sample to 68
listed South African SMEs in order to observe the effect of the debt policy using
Tobins q as a measure performance. Market-to-book value is used as a proxy for
Tobins q. The regression model can also be estimated as follows:
Tobins:q
i;t
b
0
b
1
SDC
i;t
b
2
FS
i;t
b
3
SG
i;t
m
it
5
Tobins:q
i;t
b
0
b
1
LDC
i;t
b
2
FS
i;t
b
3
SG
i;t
m
it
6
Tobins:q
i;t
b
0
b
1
TDC
i;t
b
2
SFS
i;t
b
3
SG
i;t
m
it
7
Tobins:q
i;t
b
0
b
1
TCC
i;t
b
2
FS
i;t
b
3
SG
i;t
m
it
8
where the explanatory variables are as dened previously.
3.1.1 Estimation method. The study employs Generalized Least Squares (GLS)
panel model for the estimation. Panel data involves the pooling of observations on a
cross-section of units over several time periods. Panel data approach is more useful
than either cross-section or time-series data alone. One advantage of using the panel
data set is that, because of the several data points, degrees of freedom are increased
and collinearity among the explanatory variables is reduced, thus the efciency of
economic estimates is improved. Also, panel data can control for individual
JRF
8,4
368
heterogeneity due to hidden factors, which, if neglected in time-series or cross section
estimations leads to biased results (Baltagi, 1995). The panel regression equation
differs from a regular time-series or cross-section regression by the double subscript
attached to each variable. The general form of the model can be written as:
Y
it
b
o
b
1
X
it
m
it
9
Here, m
it
is a random term and m
it
m
i
n
it
; where m
i
is the rm specic effects and
n
it
is a random term.
The choice of the model estimation whether random effects or xed effects will
depend on the underlying assumptions. In a randomeffect model, m
i
and n
it
are random
with known disturbances. In a xed effects m
i
, the rm-specic effects, and n
it
, a
random term, are xed parameters and are estimated together with the other
parameters. For most panel applications, a one-way error component model for the
disturbances is adopted, with m
it
m
i
n
it
; where m
i
accounts for any unobservable
rm-specic effects that is not included in the regression model, and n
it
represents the
remaining disturbances in the regression which varies with individual rms and time.
4. Empirical results
4.1 Descriptive summary statistics
Table I provides the descriptive statistics of all the variables used. The mean
short-term ratio, long-term debt ratio, total debt ratio, and trade credit to capital ratio
for the Ghanaian sample are shown as 0.3761, 0.0518, 0.4001, and 0.2427 respectively.
The total assets of the sampled Ghanaian SMEs are valued on the average at
7.67e 09 Ghanaian cedis. The mean growth rate in sales is 50.39 per cent. Gross
prot margin and return on assets also register average rates of 39.51 per cent and 9.25
per cent. The mean values of all the variables are signicant at 1 per cent level. With
Mean Standard error SD t-statistics p-value
Ghana
SDC 0.3761 0.0109 0.2876 34.3483 0.0000
LDC 0.0518 0.0058 0.1507 8.9855 0.0000
TDC 0.4001 0.0113 0.2985 35.4914 0.0000
TCC 0.2427 0.0095 0.2534 25.5503 0.0000
SIZE 7.67e 09 1.69e 09 4.42e 10 4.5329 0.0000
SG 0.5039 0.4588 1.0503 10.9822 0.0000
GPM 0.3951 0.1202 3.1948 3.2862 0.0011
ROA 0.0925 0.0130 0.3391 7.1236 0.0000
South Africa
SDC 0.3317 0.0112 0.2975 29.6616 0.0000
LDC 0.1874 0.0143 0.3770 13.1375 0.0000
TDC 0.4989 0.0180 0.4783 27.6529 0.0000
TCC 0.1963 0.0091 0.1999 21.6262 0.0000
SIZE 1.94e 08 2.44e 07 6.64e 08 7.9611 0.0000
SG 2.1914 1.0231 23.7519 2.1420 0.0326
GPM 21.1644 1.0798 22.2872 21.0783 0.2815
ROA 20.1862 1.3958 36.5051 20.1334 0.8939
Tobins q 11.7399 8.3907 181.324 1.3992 0.1624
Table I.
Summary statistics
Debt policy
369
respect to the South African sample, the mean debt ratios are given as 0.3317 for
short-term debt; 0.1874 for long-term debt; 0.4989 for total debt and 0.1963 for trade
credit to capital ratio. The average value of total assets is 1.94e 08 South African
rand and the average growth rate is 219.14 per cent. The mean gross prot margin and
return on assets are also indicated as 2116.44 per cent and 218.62 per cent
respectively. The average market to book value ratio or Tobins q for only the listed
South African SMEs is given as 11.7399. The mean values of short-term debt,
long-term debt, total debt ratio, and trade credit are all signicant at 1 per cent level.
The mean value of rm size is also signicant at 5 per cent level. The mean values of
the performance variables (i.e. gross prot margin, return on assets and Tobins q) are
not signicant at conventional levels.
An additional test was performed to compare the debt ratios of Ghanaian and South
African SMEs. The t-test of hypothesis of equal means had t-test values of 2.8366,
28.8091, 24.6433, and 3.5323 for short-term debt, long-term debt, total debt, and trade
credit respectively, as shown in Table II. These values are all signicant at 1 per cent
levels. The test results suggest that the null hypothesis that capital structure is the
same across the countries can be rejected. That is, capital structure varies across the
two countries. The results show that Ghanaian SMEs exhibit signicantly higher
short-term debt and trade credits than South African SMEs, whereas South African
SMEs are signicantly more likely to employ long-term debt than Ghanaian SMEs. In
terms of total debt, the results again indicate that South African SMEs have
signicantly more total debt in their capital structure than their Ghanaian
counterparts. These differences may be attributable to differences in economic
environments, nancial markets and economies of scale.
4.2 Regression results
Regression analyses are carried out to establish the relationship between capital
structure and performance. Measures of performance are regressed against different
measures of capital structure. The F-statistic and Hausman test were used to test the
validity of xed and random effects. The GLS regression was however found to be a
more robust and appropriate specication. The estimation was done using E-Views
version 5. The GLS heteroscedastic-consistent panel regression results are presented in
Tables III to VII below.
Tables III and IV show the regression results, using gross prot as a measure of
performance. The effect of short-term debt is signicantly and negatively associated
with gross prot margin for both Ghana and South Africa. This indicates that
Sample group Short-term debt ratio Long-term debt ratio Total debt ratio Trade credit ratio
Ghana 0.3761 0.0518 0.4001 0.2427
South Africa 0.3317 0.1874 0.4989 0.1963
Combined 0.3536 0.1204 0.4496 0.2239
Diff 0.0444 20.1355 20.0988 0.0464
t-statistics 2.8366
*
-8.8091
*
24.6433
*
3.5323
*
Notes:
*
Signicant at 1% level; Test: Ho: mean (Xi)
Ghana
2 mean (Xi)
South Africa
0; Ha: mean
(Xi)
Ghana
2 mean (Xi)
SouthAfrica
0 where Xi measures of capital structure
Table II.
Mean debt ratios across
sample groups
JRF
8,4
370
P
r
o

t
a
b
i
l
i
t
y
:
g
r
o
s
s
p
r
o

t
m
a
r
g
i
n
(
1
)
(
2
)
(
3
)
(
4
)
V
a
r
i
a
b
l
e
N
S
t
a
n
d
a
r
d
e
r
r
o
r
N
S
t
a
n
d
a
r
d
e
r
r
o
r
N
S
t
a
n
d
a
r
d
e
r
r
o
r
N
S
t
a
n
d
a
r
d
e
r
r
o
r
C
o
n
s
t
a
n
t
2
0
.
4
0
6
7
*
0
.
0
0
7
8
2
0
.
3
0
7
6
*
0
.
0
0
5
3
0
.
1
3
0
5
*
0
.
0
4
2
9
2
0
.
3
1
0
4
*
0
.
0
0
7
1
S
D
C
2
0
.
2
1
9
8
*
0
.
0
0
1
7
L
D
C
0
.
9
6
2
5
*
0
.
0
6
6
1
T
D
C
2
0
.
1
7
1
0
*
0
.
0
0
9
4
T
C
C
2
0
.
2
2
6
4
*
0
.
0
0
1
9
L
o
g
(
S
I
Z
E
)
0
.
0
3
8
4
*
0
.
0
0
0
4
0
.
0
2
7
2
*
0
.
0
0
0
3
0
.
0
0
9
2
*
0
.
0
0
2
1
0
.
0
3
3
0
*
0
.
0
0
0
4
L
o
g
(
S
G
)
0
.
0
1
0
9
*
0
.
0
0
0
5
0
.
0
3
1
7
*
0
.
0
0
0
3
2
0
.
0
0
9
4
*
0
.
0
0
2
4
0
.
0
2
3
3
*
0
.
0
0
0
5
R
-
s
q
u
a
r
e
d
0
.
5
8
6
2
0
.
5
2
1
7
0
.
6
8
7
5
0
.
5
4
2
5
A
d
j
u
s
t
e
d
R
-
s
q
u
a
r
e
d
0
.
5
8
3
4
0
.
5
1
8
5
0
.
6
8
5
4
0
.
5
3
9
5
F
-
s
t
a
t
i
s
t
i
c
s
2
0
8
.
2
5
4
4
*
1
5
9
.
2
6
9
3
*
3
2
9
.
2
5
3
7
*
1
8
0
.
6
4
6
3
*
N
o
t
e
s
:
*
I
n
d
i
c
a
t
e
s
i
g
n
i

c
a
n
c
e
a
t
l
e
v
e
l
s
o
f
1
%
,
s
t
a
n
d
a
r
d
e
r
r
o
r
s
i
n
p
a
r
e
n
t
h
e
s
e
s
;
(
1
)
r
e
p
r
e
s
e
n
t
s
r
e
g
r
e
s
s
i
o
n
r
e
s
u
l
t
s
f
o
r
s
h
o
r
t
-
t
e
r
m
d
e
b
t
;
(
2
)
r
e
p
r
e
s
e
n
t
s
r
e
g
r
e
s
s
i
o
n
r
e
s
u
l
t
s
f
o
r
l
o
n
g
-
t
e
r
m
d
e
b
t
;
(
3
)
r
e
p
r
e
s
e
n
t
s
r
e
g
r
e
s
s
i
o
n
r
e
s
u
l
t
s
f
o
r
t
o
t
a
l
d
e
b
t
;
(
4
)
r
e
p
r
e
s
e
n
t
s
r
e
g
r
e
s
s
i
o
n
r
e
s
u
l
t
s
f
o
r
t
r
a
d
e
c
r
e
d
i
t
Table III.
Regression results debts
on gross prot margin
(Ghana)
Debt policy
371
P
r
o

t
a
b
i
l
i
t
y
:
g
r
o
s
s
p
r
o

t
m
a
r
g
i
n
V
a
r
i
a
b
l
e
(
1
)
(
2
)
(
3
)
(
4
)
N
S
t
a
n
d
a
r
d
e
r
r
o
r
N
S
t
a
n
d
a
r
d
e
r
r
o
r
N
S
t
a
n
d
a
r
d
e
r
r
o
r
N
S
t
a
n
d
a
r
d
e
r
r
o
r
C
o
n
s
t
a
n
t
0
.
2
7
5
7
*
*
*
0
.
0
1
2
7
0
.
1
3
4
9
*
*
*
0
.
0
4
3
3
0
.
3
8
6
5
*
*
*
0
.
0
3
5
8
0
.
0
8
2
7
*
*
*
0
.
0
2
0
0
S
D
C
2
0
.
3
9
9
5
*
*
*
0
.
0
0
6
3
L
D
C
0
.
2
9
5
6
*
*
*
0
.
0
3
3
0
T
D
C
2
0
.
2
6
6
4
*
*
*
0
.
0
0
9
0
T
C
C
2
0
.
4
6
5
1
*
*
*
0
.
0
0
9
8
L
o
g
(
S
I
Z
E
)
0
.
0
0
3
6
*
*
*
0
.
0
0
0
8
0
.
0
0
0
3
0
.
0
0
2
4
2
0
.
0
0
3
7
*
0
.
0
0
2
0
0
.
0
1
2
8
*
*
*
0
.
0
0
1
2
L
o
g
(
S
G
)
0
.
0
0
2
3
*
0
.
0
0
1
3
0
.
0
0
5
9
*
*
0
.
0
0
2
7
0
.
0
0
3
4
*
*
0
.
0
0
1
7
0
.
0
0
8
3
*
*
*
0
.
0
0
1
5
R
-
s
q
u
a
r
e
d
0
.
5
9
8
3
0
.
1
8
8
0
.
4
4
5
0
0
.
5
3
5
8
A
d
j
u
s
t
e
d
R
-
s
q
u
a
r
e
d
0
.
5
9
2
6
0
.
1
7
6
8
0
.
4
3
7
2
0
.
5
2
9
3
F
-
s
t
a
t
i
s
t
i
c
s
1
0
5
.
7
3
1
9
*
*
*
1
6
.
4
6
8
5
*
*
*
5
6
.
9
3
4
7
*
*
*
5
6
.
9
3
4
7
*
*
*
N
o
t
e
s
:
*
,
*
*
,
*
*
*
i
n
d
i
c
a
t
e
s
i
g
n
i

c
a
n
c
e
a
t
l
e
v
e
l
s
o
f
1
0
%
,
5
%
a
n
d
1
%
r
e
s
p
e
c
t
i
v
e
l
y
Table IV.
Regression results debts
on gross prot margin
(South Africa)
JRF
8,4
372
P
r
o

t
a
b
i
l
i
t
y
:
r
e
t
u
r
n
o
n
a
s
s
e
t
s
V
a
r
i
a
b
l
e
(
1
)
(
2
)
(
3
)
(
4
)
N
S
t
a
n
d
a
r
d
e
r
r
o
r
N
S
t
a
n
d
a
r
d
e
r
r
o
r
N
S
t
a
n
d
a
r
d
e
r
r
o
r
N
S
t
a
n
d
a
r
d
e
r
r
o
r
C
o
n
s
t
a
n
t
0
.
3
7
0
1
*
*
0
.
0
2
6
5
0
.
3
8
5
4
*
*
0
.
0
2
4
5
0
.
3
0
0
3
*
*
0
.
0
2
2
0
0
.
3
5
3
4
*
*
0
.
0
2
4
2
S
D
C
2
0
.
1
0
2
1
*
*
0
.
0
0
6
3
L
D
C
2
0
.
1
2
6
0
*
*
0
.
0
1
0
2
T
D
C
2
0
.
1
0
0
3
*
*
0
.
0
0
4
4
T
C
C
2
0
.
0
8
5
4
*
*
0
.
0
0
4
9
L
o
g
(
S
I
Z
E
)
2
0
.
0
1
0
9
*
*
0
.
0
0
1
3
2
0
.
0
1
3
2
*
*
0
.
0
0
1
1
2
0
.
0
0
7
6
*
*
0
.
0
0
1
1
2
0
.
0
1
1
2
*
*
0
.
0
0
1
1
L
o
g
(
S
G
)
0
.
0
0
1
8
0
.
0
0
1
6
0
.
0
0
5
5
*
*
0
.
0
0
1
7
0
.
0
0
3
8
*
*
0
.
0
0
1
1
0
.
0
0
3
0
*
0
.
0
0
1
4
R
-
s
q
u
a
r
e
d
0
.
3
3
8
4
0
.
3
1
0
4
0
.
3
3
4
6
0
.
3
2
5
8
A
d
j
u
s
t
e
d
R
-
s
q
u
a
r
e
d
0
.
3
3
3
9
0
.
3
0
5
7
0
.
3
3
0
1
0
.
3
2
1
4
F
-
s
t
a
t
i
s
t
i
c
s
7
5
.
1
7
9
4
*
*
6
5
.
7
2
2
1
*
*
7
5
.
2
5
5
3
*
*
7
3
.
6
1
1
5
N
o
t
e
s
:
*
,
*
*
i
n
d
i
c
a
t
e
s
i
g
n
i

c
a
n
c
e
a
t
l
e
v
e
l
s
o
f
5
%
a
n
d
1
%
r
e
s
p
e
c
t
i
v
e
l
y
Table V.
Regression results debts
on return on assets
(Ghana)
Debt policy
373
P
r
o

t
a
b
i
l
i
t
y
:
r
e
t
u
r
n
o
n
a
s
s
e
t
s
V
a
r
i
a
b
l
e
(
1
)
(
2
)
(
3
)
(
4
)
N
S
t
a
n
d
a
r
d
e
r
r
o
r
N
S
t
a
n
d
a
r
d
e
r
r
o
r
N
S
t
a
n
d
a
r
d
e
r
r
o
r
N
S
t
a
n
d
a
r
d
e
r
r
o
r
C
o
n
s
t
a
n
t
1
.
7
7
8
0
*
*
0
.
0
1
1
1
1
.
2
1
0
8
*
*
0
.
0
1
9
3
0
.
9
5
7
1
*
*
0
.
0
2
0
4
2
0
.
1
1
5
2
*
*
0
.
0
1
4
5
S
D
C
0
.
3
0
0
3
*
*
0
.
0
0
1
7
L
D
C
2
0
.
3
9
9
4
*
*
0
.
0
0
7
6
T
D
C
2
0
.
0
3
7
4
*
*
0
.
0
0
2
1
T
C
C
0
.
0
2
3
6
*
0
.
0
0
9
4
L
o
g
(
S
I
Z
E
)
2
0
.
1
0
8
2
*
*
0
.
0
0
0
6
2
0
.
0
5
7
9
*
*
0
.
0
0
1
0
2
0
.
0
4
7
5
*
*
0
.
0
0
1
1
0
.
0
1
1
6
*
*
0
.
0
0
0
7
L
o
g
(
S
G
)
2
0
.
0
5
0
9
*
*
0
.
0
0
0
1
0
.
0
0
9
8
*
*
0
.
0
0
0
2
2
0
.
0
0
9
4
*
*
0
.
0
0
0
2
0
.
0
0
5
6
*
*
0
.
0
0
0
9
R
-
s
q
u
a
r
e
d
0
.
6
5
5
7
0
.
1
5
9
7
0
.
0
6
0
1
0
.
5
1
2
8
A
d
j
u
s
t
e
d
R
-
s
q
u
a
r
e
d
0
.
6
5
2
5
0
.
1
5
1
8
0
.
0
5
1
3
0
.
5
0
6
0
F
-
s
t
a
t
i
s
t
i
c
s
2
0
4
.
3
8
8
7
*
*
2
0
.
0
8
9
3
*
*
6
.
8
0
5
0
*
*
7
4
.
7
4
2
9
*
*
N
o
t
e
s
:
*
,
*
*
i
n
d
i
c
a
t
e
s
i
g
n
i

c
a
n
c
e
a
t
l
e
v
e
l
s
o
f
5
%
a
n
d
1
%
r
e
s
p
e
c
t
i
v
e
l
y
Table VI.
Regression results debts
on return on assets (South
Africa)
JRF
8,4
374
T
o
b
i
n

s
q
V
a
r
i
a
b
l
e
(
5
)
(
6
)
(
7
)
(
8
)
N
S
t
a
n
d
a
r
d
e
r
r
o
r
N
S
t
a
n
d
a
r
d
e
r
r
o
r
N
S
t
a
n
d
a
r
d
e
r
r
o
r
N
S
t
a
n
d
a
r
d
e
r
r
o
r
C
o
n
s
t
a
n
t
0
.
2
2
7
6
*
0
.
1
2
1
0
2
0
.
0
4
5
6
0
.
1
6
1
7
0
.
2
7
4
5
0
.
1
6
9
3
0
.
1
4
0
7
0
.
1
0
3
3
S
D
C
0
.
6
3
9
3
*
*
0
.
0
4
5
0
L
D
C
2
0
.
9
6
9
5
*
*
0
.
0
9
8
1
T
D
C
2
0
.
6
5
1
7
*
*
0
.
0
5
3
4
T
C
C
1
.
1
3
3
2
*
*
0
.
0
4
6
8
L
o
g
(
S
I
Z
E
)
0
.
0
5
1
0
*
*
0
.
0
0
7
0
0
.
0
8
5
0
*
*
0
.
0
0
8
5
0
.
0
7
3
7
*
*
0
.
0
0
9
6
0
.
0
5
2
8
*
*
0
.
0
0
5
9
L
o
g
(
S
G
)
0
.
1
5
8
4
*
*
0
.
0
0
8
2
0
.
0
9
7
2
*
*
0
.
0
1
1
0
0
.
1
3
7
6
*
*
0
.
0
0
9
4
0
.
1
5
7
5
*
*
0
.
0
0
8
4
R
-
s
q
u
a
r
e
d
0
.
5
6
2
0
0
.
5
4
8
0
0
.
4
0
1
8
0
.
5
5
2
7
A
d
j
u
s
t
e
d
R
-
s
q
u
a
r
e
d
0
.
5
5
5
6
0
.
5
4
1
2
0
.
3
9
2
9
0
.
5
4
6
0
F
-
s
t
a
t
i
s
t
i
c
s
8
5
.
9
8
3
8
*
*
8
1
.
2
2
7
6
*
*
4
5
.
0
0
1
8
*
*
8
2
.
7
8
8
5
*
*
N
o
t
e
s
:
*
,
*
*
i
n
d
i
c
a
t
e
s
i
g
n
i

c
a
n
c
e
a
t
l
e
v
e
l
s
o
f
1
0
%
a
n
d
1
%
r
e
s
p
e
c
t
i
v
e
l
y
;
(
5
)
r
e
p
r
e
s
e
n
t
s
r
e
g
r
e
s
s
i
o
n
r
e
s
u
l
t
s
f
o
r
s
h
o
r
t
-
t
e
r
m
d
e
b
t
;
(
6
)
r
e
p
r
e
s
e
n
t
s
r
e
g
r
e
s
s
i
o
n
r
e
s
u
l
t
s
f
o
r
l
o
n
g
-
t
e
r
m
d
e
b
t
;
(
7
)
r
e
p
r
e
s
e
n
t
s
r
e
g
r
e
s
s
i
o
n
r
e
s
u
l
t
s
f
o
r
t
o
t
a
l
d
e
b
t
;
(
8
)
r
e
p
r
e
s
e
n
t
s
r
e
g
r
e
s
s
i
o
n
r
e
s
u
l
t
s
f
o
r
t
r
a
d
e
c
r
e
d
i
t
Table VII.
Regression results debts
on Tobins q (South
Africa)
Debt policy
375
increasing the amount of short-term debt will result in a decrease in the gross prot
margin of the rms. The results also show that long-term debt has a signicantly
positive relationship with gross prot margin for both countries. SMEs that employ
more long-term debt record higher gross prot margin. The relation between total debt
to capital ratio and gross prot margin was found to be signicant and negative for
both countries. In terms of trade credit, the results indicate a statistically signicant
and negative association between trade credit and gross prot margin for both Ghana
and South Africa. Increasing trade credit in the rms capital structure is associated
with decreasing gross prot margin. In the Ghanaian sample, the control variables
(size and sales growth) reveal statistically signicant positive effects on gross prot
margin for all measures of debt with the exception of the total debt measure, where
growth is shown to have signicantly negative relation with gross prot margin. Also,
in the South African sample, size indicates positive relations with gross prot margin
for measures of short-term debt and trade credit. The result for total debt is negative
and that of long-term debt is insignicant. But sales growth indicates a signicantly
positive relation with gross prot margin for all measures of debt.
Tables V and VI report the regression results using return on assets as the
performance measure. In the case of Ghana, the results as shown in Table V exhibit
signicantly negative relations between all the measures of capital structure and
return on assets. For Ghanaian SMEs, adopting a high debt policy is signicantly more
likely to lead to lower protability (return on assets). Increasing the proportion of debt
in the rms capital structure could result in high bankruptcy cost and this is likely to
impact negatively on return on assets. Also, the results provide support for the
argument that due to agency conicts, SMEs over-leverage themselves, thus
negatively affecting their own performance. This is also consistent with the ndings of
Gleason et al. (2000). In the Ghanaian sample, rm size indicates signicantly negative
relation with return on assets for all the measures of debt. Sales growth also shows
statistically positive relation with long-term debt, total debt, and trade credit. The
relationship between sales growth and return on assets for short-term debt is not
statistically signicant.
In the South African sample, the results as illustrated in Table VI reveal a
statistically signicant positive relationship between short-term debt and return on
assets. Similarly, the results indicate a statistically signicant relationship between
trade credit and return on assets. This might be attributed to the fact that short-term
debt and trade credit seem to be relatively less costly. Therefore, increasing short-term
debt or trade credit with relatively low interest rate could result in high prot levels.
The regression results show signicantly negative association between return on
assets and long-term debt, and total debt. This also suggests that long-term debt
attracts higher cost and therefore employing high proportions of long-term debt in the
SMEs capital structure could lead to low return on assets. The results from the South
African data imply that pursuing a high long-term debt strategy might be associated
with lowprotability. This position supports the ndings of previous empirical studies
(see Fama and French, 1998; Abor, 2005). The results from the South African data also
reveal signicantly negative interaction between rm size and return on assets for
measures of short-term debt, long-term debt, and total debt but a statistically
signicant positive association between size and return on assets for the trade credit
model. The sales growth variable exhibits signicantly negative effect on return on
JRF
8,4
376
assets for measures of short-term debt and total debt but statistically signicant
positive impact on return on assets for measures of long-term debt, and trade credit.
The analysis was also done considering only listed SMEs in South Africa. The
essence of this was to examine the effects of the various measures of capital structure
on Tobins q. The results as shown in Table VII show statistically signicant positive
relationships between Tobins q and two measures of capital structure (short-term debt
and trade credit). The results, however, indicate signicantly negative relation between
the Tobins q and long-term debt, and total debt ratios. In other words, increasing the
amount of short-term debt and trade credit in the rms debt structure is signicantly
more likely to positively inuence their Tobins q or market-to-book value. Also, a rise
in the long-term debt and total debt implies a reduction in the Tobins q. The results of
this study suggest that for listed SMEs employing more short-term debt and trade
credit has the tendency of causing an improvement in their market-to-book value but
having more long-term debt would lead to a negative impact on the market-to-book
value of the rms. The results show that large rm size and high sales growth are
associated with improvement in the Tobins q or market-to-book value.
5. Conclusions and implications
One important nancial decision rms are confronted with is the debt policy or capital
structure choice. This decision is particularly crucial given the effect it has on the value
of the rm. This study has examined the relationship between capital structure and
performance of SMEs in Ghana and South Africa during a six-year period, 1998-2003.
The empirical results indicate that short-term debt is signicantly and negatively
related to gross prot margin for both Ghana and South Africa. The results show that
long-term debt has a signicantly positive relationship with gross prot margin for
both countries. The relation between total debt ratio and gross prot margin was found
to be signicant and negative. The results also reveal a statistically signicant and
negative association between trade credit and gross prot margin for both Ghana and
South Africa. In the case of Ghana, the results show signicantly negative relations
between all the measures of capital structure and return on assets. In the South African
sample, the results reveal signicantly positive relationships between return on assets
and short-term debt, and trade credit. However in terms of long-term debt and total
debt, the results show statistically signicant negative relationship between return on
assets and both long-term debt and total debt. The results of this paper also show, for
the listed SMEs, statistically signicant positive relationship between Tobins q and
two measures of capital structure (short-term debt and trade credit) but indicate
signicantly negative relations between the Tobins q and long-term debt, and total
debt ratio.
The results of this study have shown that in the presence of control variables,
capital structure has a signicant inuence on the performance of SMEs. By and large,
the results indicate that capital structure, especially long-term and total debt ratios
negatively affect performance of SMEs. The negative relationships imply that SMEs
generally are averse to use more equity because of the fear of losing control and
therefore employ more debt in their capital structure than would be appropriate. Apart
from the problems SMEs face in acquiring equity, one reason for increasing debt use
may be to avoid agency conicts. Employing debt excessively is likely to result in high
bankruptcy cost which could negatively affect performance. SMEs that pursue very
Debt policy
377
high debt policy compared to the industry average should also consider increasing the
equity component in their capital structure in order to avoid the negative effects of
excessive debt on performance.
References
Abor, J. (2005), The effect of capital structure on protability: empirical analysis of listed rms
in Ghana, Journal of Risk Finance, Vol. 6 No. 5, pp. 438-45.
Baker, S.H. (1973), Risk, leverage and protability: an industry analysis, Review of Economics
and Statistics, Vol. 55, pp. 503-7.
Baltagi, B.H. (1995), Econometric Analysis of Panel Data, Wiley, Chichester.
Chaganti, R., DeCarolis, D. and Deeds, D. (1995), Predictors of capital structure in small
ventures, Entrepreneurship Theory and Practice, Winter, pp. 1042-2587.
Champion, D. (1999), Finance: the joy of leverage, Harvard Business Review, Vol. 77 No. 4,
pp. 19-22.
Fama, E.F. and French, K.R. (1998), Taxes, nancing decisions, and rm value, The Journal of
Finance, Vol. 53, pp. 819-43.
Gleason, K.C., Mathur, L.K. and Mathur, I. (2000), The interrelationship between culture, capital
structure, and performance: evidence from European retailers, Journal of Business
Research, Vol. 50, pp. 185-91.
Grabowski, H.G. and Mueller, R. (1972), Managerial and stockholder welfare models of rm
expenditures, Review of Economics and Statistics, Vol. 54, pp. 9-24.
Graham, J.R. and Harvey, C. (2001), The theory and practice of corporate nance: evidence from
the eld, Journal of Financial Economics, Vol. 60, pp. 187-243.
Hadlock, C.J. and James, C.M. (2002), Do banks provide nancial slack?, The Journal of Finance,
Vol. 57, pp. 1383-420.
Hammes, K. (1998), Various aspects of capital structure in Poland, working paper, Tallin
Technical University, Tallin.
Hammes, K. (2003), Firm performance, debt, bank loans and trade credit. An empirical study,
working paper, Department of Economics, Gothenburg University, Gothenburg.
Heinkel, R. (1982), A theory of capital structure relevance under imperfect information, Journal
of Finance, Vol. 37, pp. 1141-50.
Hutchinson, R.W. (1995), The capital structure and investment decisions of the small
owner-managed rm: some explanatory issues, Small Business Economics, Vol. 7, p. 231.
Jensen, M. (1986), Agency cost of free cash-ow, corporate nance and takeovers, American
Economic Review, Vol. 76, pp. 323-9.
Jensen, M. and Meckling, W. (1976), Theory of the rm: managerial behaviour, agency costs and
ownership structure, Journal of Financial Economics, Vol. 3, pp. 305-60.
Kang, J. and Schivdasani, A. (1995), Firm performance, corporate governance and top executive
turnover in Japan, Journal of Financial Economics, Vol. 38, pp. 29-58.
Kaplan, S. and Minton, B. (1994), Appointment of outsiders to japanese boards: determinants
and implications for managers, Journal of Financial Economics, Vol. 36, pp. 225-57.
Krishman, S.V. and Moyer, C.R. (1997), Performance, capital structure and home country: an
analysis of Asian corporations, Global Finance Journal, Vol. 8, pp. 129-43.
JRF
8,4
378
Majumdar, S.K. and Chhibber, P. (1999), Capital structure and performance: evidence from a
transition economy on an aspect of corporate governance, Public Choice, Vol. 98,
pp. 287-305.
Mesquita, J.M.C. and Lara, J.E. (2003), Capital structure and protability: the Brazilian case,
working paper, Academy of Business and Administration Sciences Conference,
Vancouver, July 11-13.
Miller, M.H. and Rock, K. (1985), Dividend policy under asymmetric information, Journal of
Finance, Vol. 40, pp. 1031-51.
Modigliani, F. and Miller, M. (1958), The cost of capital, corporation nance and the theory of
investment, The American Economic Review, Vol. 48 No. 3, pp. 261-97.
Modigliani, F. and Miller, M. (1963), Corporate income taxes and the cost of capital: a
correction, American Economic Review, Vol. 53, pp. 443-53.
Myers, S.C. (1984), The capital structure puzzle, Journal of Finance, Vol. 39, pp. 575-92.
Myers, S.C. and Majluf, N.S. (1984), Corporate nancing and investment decisions when rms
have information that investors do not have, Journal of Financial Economics, Vol. 13,
pp. 187-221.
Nerlove, M. (1968), Factors affecting differences among rates of return on investments in
individual common stocks, Review of Economics and Statistics, Vol. 50, pp. 312-31.
Noe, T. (1988), Capital structure and signaling game equilibria, Review of Financial Studies,
Vol. 1, pp. 331-55.
Petersen, M.A. and Rajan, R.G. (1994), The benets of lending relationships: evidence from
small business data, The Journal of Finance, Vol. 49, pp. 3-37.
Roden, D.M. and Lewellen, W.G. (1995), Corporate capital structure decisions: evidence from
leveraged buyouts, Financial Management, Vol. 24, pp. 76-87.
Ross, S. (1977), The determination of nancial structure: the incentive signalling approach, Bell
Journal of Economics, Vol. 8, pp. 23-40.
Smith, C. (1986), Investment banking and the capital acquisition process, Journal of Financial
Economics, Vol. 15, pp. 3-29.
Taub, A.J. (1975), Determinants of the rms capital structure, The Review of Economics and
Statistics, Vol. 57, pp. 410-6.
Titman, S. (1984), The effect of capital structure on a rms liquidation decisions, Journal of
Financial Economics, Vol. 13, pp. 137-51.
Further reading
Abor, J. and Biekpe, N. (2005), What determines the capital structure of listed rms in Ghana?,
African Finance Journal, Vol. 7 No. 1, pp. 37-48.
Corresponding author
Joshua Abor can be contacted at: joshabor@ug.edu.gh
Debt policy
379
To purchase reprints of this article please e-mail: reprints@emeraldinsight.com
Or visit our web site for further details: www.emeraldinsight.com/reprints
Foreign exchange risk exposure
of listed companies in Ghana
Zubeiru Salifu, Ko A. Osei and Charles K.D. Adjasi
Department of Finance, University of Ghana Business School, Ghana
Abstract
Purpose The purpose of this research is to examine the foreign exchange exposure of listed
companies on the Ghana Stock Exchange over the period January 1999 to December 2004. The
research uses different exchange rate measures namely; the cedi to US dollar, the cedi to UK pound
sterling, the cedi to the euro and a trade-weighted exchange rate index to determine the degree of
exposure.
Design/methodology/approach The Jorion (1990) two-factor model which regresses the return
on a rm against changes in the exchange rate and return on the market is used to estimate the
exchange rate exposure for the sample of twenty rms used in this study.
Findings About 55 per cent of rms in the sample have a statistically signicant exposure to the
US dollar whilst 35 per cent are statistically exposed to the UK pound sterling. Sector specic exposure
results show that the manufacturing and retail sectors are signicantly exposed to the US dollar
exchange rate risk. The nancial sector did not show any risk exposure to any of the international
currencies. The most dominant source of exchange rate risk exposure is the US dollar. Most rms are
also negatively exposed to the cedi to US dollar exchange rate changes, implying that the cedi
depreciation vis-a` -vis the US dollar adversely affects rm returns.
Originality/value The study reveals the extent of foreign exchange exposure of rms in Ghana
and also adds to the limited body of empirical literature on exchange rate exposure of rms in Africa.
Results of this study serve as a useful guide to corporate managers and investors on the degree of
foreign exchange exposure and the need to effectively manage rm exposure.
Keywords Exchange rates, Stock returns, Stock market, Ghana
Paper type Research paper
Introduction
The adoption of a oating exchange rate regime, the rapid globalization of national
economies and the attempts by multinationals to seek investment opportunities and
markets beyond their immediate borders account for the increasing exposure of rms
to foreign exchange risk in recent times. The exchange rate exposure of a rm refers to
the sensitivity of its economic value, or stock price, to exchange rate changes
(Heckman, 1983) or its economic exposure to exchange rate risk (Adler and Dumas,
1984). The frequent changes in foreign exchange rates is of prime concern to rms,
nancial analysts and economists due to its effect on rm operations, revenue and
valuation. Exchange rate variability is a source of cash ow risk for rms with
foreign-denominated assets and liabilities (whether exporters or importers), as well as
rms with overseas operations.
Additionally, rms without foreign revenues, costs, or operations might be
indirectly affected by exchange rate changes through its impact on foreign competition
or broader macro-economic conditions (Parsley and Popper, 2002). Given the difculty
in anticipating exchange rate movements, corporate managers, investors and analysts
have to devise means of managing or investing optimally to neutralise exchange rate
risks.
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1526-5943.htm
JRF
8,4
380
The Journal of Risk Finance
Vol. 8 No. 4, 2007
pp. 380-393
qEmerald Group Publishing Limited
1526-5943
DOI 10.1108/15265940710777324
Due to the adverse effects of exchange rate exposure, and the difculty in
anticipating exchange rate movements, there have been increased attempts by
researchers to determine the level and effect of exchange rate exposure of rms. A
number of studies (see for instance Adler and Dumas, 1984; Choi et al., 1992; Bodnar
and Gentry, 1993; Choi and Prasad, 1995; Chamberlain et al., 1997) have been
conducted to determine the exposure of rms to exchange rate movements. Most of
these studies have however been inconclusive on the nature of exchange rate exposure.
Again these studies have been largely conducted on industrialized economies. Even
when developing economies are considered, Sub-Saharan African (SSA) economies
remain absent in such samples. It is however important to note that rms in SSA may
also be signicantly exposed to exchange rate movements. Indeed since the 1990s,
most SSA economies have undergone economic stabilization reforms, comprising
largely of exchange rate liberalization from xed to oating regimes. The liberalization
of exchange rates and the subsequent rapid changes in exchange rates in SSA certainly
have implications for African rms. It is therefore important to determine the exposure
of such rms to exchange rate movements.
This study examines the exchange rate exposure of rms in a SSA economy-Ghana.
Ghana was the rst African country, to embark on economic stabilization and
adjustment policies within the World Bank and International Monetary Fund (IMF)
framework, in the early 1980s (Aryeetey and Tarp, 2000). Exchange rate movement in
Ghana since then has been variable with periods of rapid depreciation of the domestic
currency which, adversely affected the Ghanaian economy (Oduro and Harrigan, 2000).
Even though studies have been conducted on the exchange rate regimes and the
implications for macroeconomic management as well as managing foreign exchange
risk (Abor, 2005), there is no study on the rm exposure to exchange risk in Ghana.
The study is signicant in three main ways; rst it adds to the body of empirical
literature on exchange rate exposure of rms; second it explains the exchange rate
exposure of rms from an African perspective and thirdly, the study incorporates the
effect of different currency exchange rates including the cedi[1] to the US dollar, the
cedi to the UK pound and the cedi to the euro as well as a trade weighted exchange rate
The rest of the study is structured as follows. Section two gives a historical
overview of exchange rate developments in Ghana; section three reviews the extant
literature on foreign currency exposure. Section four presents the methodology
adopted for the study and also discusses the ndings. Conclusions from the study are
drawn in section ve.
An overview of exchange rate developments in Ghana
Ghanas exchange rate system has had a chequered history of xed and oating rate
regimes. The 1970s witnessed a largely xed exchange rate regime. However, by the
late 1970s, problems of compatibility between the xed exchange rate and the
macroeconomic policy stance emerged as the domestic ination rate accelerated above
those of her major trading partners. This subsequently triggered an appreciation in the
real exchange rate. Access to foreign exchange was restricted as the demand for it far
exceeded the supply. This therefore called for the introduction of tight import controls
by the economic managers.
The trade and exchange controls subsequently encouraged the development of a
parallel market for foreign exchange and a large black market premium began to
Foreign
exchange risk
exposure
381
emerge in 1975. Oduro and Harrigan (2000) mention four major features that
characterized the xed exchange rate regime. These include a highly overvalued
ofcial exchange rate; an active parallel market in foreign exchange; capital controls as
well as an allocation of foreign currency based on import licenses. This period was also
characterized by an extreme hesitance on the part of the monetary and scal
authorities to implement large exchange rate adjustments for fear of worsening the
political instability at that time.
As part of a broader donor supported Economic Recovery Programme (ERP)
introduced in 1983, Ghana moved from a xed exchange rate regime towards a oating
exchange rate mechanism. In the period between 1983 and 1990, the government
implemented a wide range of trade and payments policies with the objective of
switching away from direct government intervention and controls towards increased
reliance on market determined outcomes. Dordunoo (1994) mentions four major
objectives of reforms to the foreign exchange regime which were a key part of the ERP.
These objectives included realigning of the ofcial rates; achieving a convergence of
ofcial and parallel rates; absorbing the parallel market into the legal market and
nally, allowing the forces of demand and supply to determine the rate and allocation
of foreign exchange. By the mid-1990s, the exchange rate regime had become virtually
fully liberalized. The local currency, (the cedi) has however experienced some huge
depreciation against the major foreign currencies since then. These rapid changes in
exchange rate movements had obvious implications for rms. As a way of reducing
such variability, the Bank of Ghana was quite instrumental in managing the exchange
rate through market forces to ensure stability.
Review of relevant literature
Foreign exchange exposure is dened by Adler and Dumas(1984) as the degree to
which the value of a rm is affected by changes in exchange rate. The literature
generally identies four channels of foreign exchange exposure: translational
exposure, transactional exposure, operational exposure and contingency exposure.
Translation exposure occurs through currency mismatch and it is also related to assets
or income derived from offshore enterprise (Glaum, 1990; Grant and Soenen, 1991;
Madura, 2003). Transactional exposure arises from future cash ows such as trade
contracts and also occurs where the value of existing obligations are affected by
changes in foreign exchange rates. Operational exposure occurs where the market
position of a rm changes as a result of the effect of exchange rate changes on
competition, prices and demand. Contingency exposure occurs from possible
revaluations arising from future liabilities. The total or economic exposure of a rm
refers to all exchange rate effects through all four channels.
The foreign exchange risk exposure literature is replete with a plethora of studies
regarding the exposure of rms and countries to frequent changes in the exchange
rates of domestic currencies in relation to the currencies of trading partner
countries. Studies such as (Jorion, 1990 and 1991; Loudon, 1993; Khoo, 1994; Booth
and Rotenberg, 1990; Choi and Prasad, 1995; Bodnar and Gentry, 1993; Fang and
Loo, 1994) to mention a few have focused on the USA and the major European
stock markets. These studies have documented evidence of exposure by many of the
rms with signicant assets and cash ow denominated in foreign currencies. In
many of these studies, exposure is measured by estimating the sensitivity of stock
JRF
8,4
382
returns to exchange rate changes (Adler and Dumas, 1984; Jorion, 1990; Choi et al.,
1992; Bodnar and Gentry, 1993; Bartov and Bodnar, 1994; Choi and Prasad, 1995;
Chamberlain et al., 1997; Choi and Elyasiani, 1997; Chow et al., 1997a, 1997b; He and
Ng, 1998).
Whilst there have been many empirical studies which have examined the
relationship between foreign exchange exposure and rm value, their results have
however been mixed. Jorion, 1990, 1991; Bartov and Bodnar, 1994; and Choi and
Prasad, 1995, using USA data and Loudon (1993), and Khoo (1994) using Australian
data nd rm values to be insensitive to exchange rate movements. Choi et al. (1992)
also examined USA banks and found about 20 per cent to be signicantly exposed to
exchange rate risk over the 1975-1987 period.
Other studies such as Booth and Rotenberg (1990) who used Canadian data and
He and Ng (1998) who employed Japanese data found some evidence that foreign
exchange risk affects rm value. Out of the 171 Japanese rms they examined, they
found that approximately 25 per cent were signicantly exposed. On their part,
Bodnar and Gentry (1993) considered the degree of exposure of about 39 industries
and report that 11 out of the 39 industries they considered exhibited signicant
exposure during the period 1979-1988. The study of Choi and Prasad (1995) on the
other hand revealed 15 per cent of 409 multinational companies (MNCs) to have
signicant exposure.
In addition to the failure of most studies to establish signicant exposure of rm
value to changes in foreign exchange rate, the extant literature on foreign exchange
exposure also shows conicting results regarding a number of areas such as the time
horizon, the sample selection procedures as well as the portfolio construction. These
contrasting positions have contributed largely to ambiguous evidence on exchange
rate exposure. Bodnar and Wong (2003) have also demonstrated that different
constructions of the market portfolio give different exposures to exchange rates.
Therefore, the choice of the market portfolio in the exposure model has a substantial
impact on the resulting exposure estimates. For example, a portfolio constructed from a
value-weighted market portfolio is dominated by large rms which tend to be
multinationals and export-oriented companies. On the other hand, equal-weighted
market portfolio is usually dominated by small rms which tend to be domestic and
import-oriented. Thus, value-weighted market portfolio generates a more negative
market exposure whereas equal-weighted market portfolio generates a more positive
market exposure.
In spite of the disagreements regarding the time horizon, the sample selection
procedures and the portfolio construction, there appears to be a common
understanding on the choice of the exchange rate in almost all currency exposure
studies. With the exception of a few studies; Miller and Reuer (1998) and Dominguez
and Tesar (2001) who use multiple exchange rates instead of a single currency proxy,
many studies on currency exposure use a trade-weighted index (single currency proxy)
to measure exchange rate.
Apart from the study by Abor (2005), which establishes that most Ghanaian rms
hardly manage their foreign exchange risk in a sophisticated manner, there is very
little evidence of studies on foreign exchange risk management by rms in SSA. Most
of the empirical studies directly or indirectly related to foreign currency exposure have
tended to focus on price and exchange rate dynamics; monetary and exchange rate
Foreign
exchange risk
exposure
383
policy and price determination; exchange rate regimes and ination; the role of
exchange rate and monetary policy in managing balance of payments; institutional
reforms and management of exchange rate policy; the effect of exchange rate policy on
agricultural competitiveness; real exchange rate movements and export growth; and
the possibilities for convergence in exchange rates (see Ndunguu, 1997; Ndunguu,
1999; Atta et al., 1999; Rutasitara, 2004; Silumbu, 1995; Odubogun,1995; Amin, 1996;
Ogun, 1998).
Methodology
Data
The data for this study involved 20 rms listed on the Ghana Stock Exchange. The
selection of the 20 listed rms was based on the fact that these companies provided
enough data for the study period. Those companies which listed before 1999 met the
data requirement criteria. Monthly share price information for the period January 1999
to December 2004 required for the study were obtained from the Ghana Stock
Exchange (GSE). The currencies and for that matter the exchange rates chosen for the
study were those widely used in international transactions in Ghana. The exchange
rates were computed as the ratio of the number of local currency which purchased a
unit of foreign currency. These were the cedi to US dollar, the cedi to euro and the cedi
to UK pound. These exchange rates were obtained from the Bank of Ghana.
In addition, a trade weighted index, the Nominal Effective Exchange Rate (NEER)
was also used for the study. The NEER is dened by the IMF to represent the ratio of
an index of the period average exchange rate of the currency in question (in this
instance the cedi) to a trade weighted geometric average of exchange rates for the
currencies of selected major trading partners. The NEER is obtained from the IMFs
International Financial Statistics. The rationale for using the NEER is to determine
whether the magnitude or degree of exposure of a rm varies with the incorporation of
the trade weighted element.
Model
The theoretical framework for the exchange rate exposure of rms is based on the fact
that, exchange rate exposure has potentially positive or negative impact on the
protability and value of the rm. This is captured in the valuation process in terms of
the rms stock returns. Thus, the approach to modeling the exchange rate exposure
has been to regress the exchange rate on rms returns.
In this study, the model adopted is that of Jorion (1990). The motivation for adopting
this model is due to the fact that it incorporates the market exposure which controls for
the markets own exchange rate exposure. Thus the exchange rate risk in this model is
the residual risk faced by the rm after controlling for the market risk, Bodnar and
Wong (2000). Bodnar and Wong (2000) also show that this is the most applied and
preferred model by researchers. The model postulates the following relationship
among the return on a rm, exchange rate changes and return on the overall market:
R
it
a
i
b
i1
CUSD b
i2
CUKP b
i3
CEURO d
i
R
mt
1
it
1
R
it
a
i
b
i
NEER
t
d
i
R
mt
1
it
2
JRF
8,4
384
where:
a
i
the constant term;
R
it
the stock return for rm i at period t;
VNEER the percentage variation in Nominal Effective Exchange Rate (NEER);
R
mt
the return on the market;
CUSD the percentage change of the US dollar relative to the cedi;
CUKP the percentage change of the UK Pound Sterling relative to the cedi; and
CEURO the percentage change of the Euro in relation to the cedi.
b
i
and d
i
represent the sensitivity of company is returns to exchange rate movements
and the returns on the market respectively. A positive b
i
indicates that a rms stock
returns appreciate with a depreciation in the value of the domestic currency, whilst a
negative b
i
denotes that stock returns fall with a depreciation in value of the domestic
currency.
The returns[2] for rm i for each period t was computed as equation 3:
R
it

P
t
2P
t21
P
t21
3
where:
R
it
the holding period return in time t of asset i;
P
t
the share price of rm i at the end of period t (as at the rst day of the
second; and subsequent months of trading); and
P
t21
the share price of rm i at the end of period t 2 1 (as at the rst day of the
rst trading month).
Equations (1) and (2) are run separately for each of the 20 rms using monthly data for
the period January 1999 to December 2004. A portfolio of sector returns is then
constructed and the process repeated for each sector portfolio. The motivation for
constructing a portfolio is due to the fact that exchange rate exposure of an industry
may vary according to its characteristics. The delineation of the sample of rms into
the three broad sector designations reects the classications by the Ghana Stock
Exchange (GSE).
The manufacturing sector (11 rms) is composed largely of some of Ghanas large
sized rms. Examples of the rms in this sector include the brewery rms,
confectionary (food), aluminium and plastics, as well as rms that produce home care
products. Firms under this sector are mostly subsidiaries of large multinational rms
and hence have a substantial value of their assets and liabilities denominated in the
currencies of their parent companies. The rms in this sector also depend substantially
on foreign imported raw materials and machinery. The retail sector (4 rms) is mainly
made up of the automobile and oil marketing companies. Firms in this group are also
large and are subsidiaries of multinational companies. Finally, the nancial sector (5
Foreign
exchange risk
exposure
385
rms) comprises mainly the commercial banks, insurance companies and other
nancial service providers listed on the exchange.
Empirical results
The results shown in Table I indicate regression results based on equation (1) on rms
exposure to multiple exchange rates; the cedi to US dollar, the cedi to euro[3] and the
cedi to the UK pound sterling. An important feature of the results in this table is that,
the US dollar is the most dominant source of foreign exchange risk. Out of the twenty
rms, eleven representing 55 per cent had signicant exposures to the US dollar
exchange rate risk. This makes the US dollar the most dominant source of exchange
risk among the major currencies of international trade and investment. This may not
be far fetched since the US dollar is the main currency for Ghanas international
transactions.
In the case of the UK pound, seven rms representing 35 per cent of the rms had
signicant exposure to the UK pound exchange rate risk whilst for the euro[4], only
ve listed rms were signicantly exposed. It is worth noting that all rms
signicantly exposed to the UK pound exchange rate risk were also signicantly
exposed to the US dollar exchange rate risk.
Almost all (ten) of the eleven rms with signicant exposure to the US dollar
exchange rate risk, experienced negative exposures. This implies that the stock returns
of these rms fell as the cedi depreciated against the US dollar. Thus these rms are
adversely affected by their exposure to the US dollar exchange rate risk. On the other
hand, all seven rms with signicant exposure to the UK pound had positive exposure
coefcients, implying that stock returns appreciated with depreciation in the value of
the cedi relative to the UK pound. It therefore appears that rms which had majority of
their transactions in UK pound beneted by using this currency in their international
transactions.
Table II displays regression results of equation (2) on exposure to the trade
weighted index (NEER). In this analysis, only seven of the 20 rms have signicant
exposure to the NEER. Of this number, four have a positive exposure whilst three have
negative exposure. This seems to suggest that the rms may not be exposed equally to
all the currencies in the index and to the trade weights used in the index. Even if a rm
is exposed to all the currencies in the index, the relative importance of the currencies in
the index as measured by their weights may not correspond exactly to that of the rm.
The use of the trade weighted index appears therefore to underestimate the extent of
foreign exchange exposure.
Table III shows the regression results of equation (1) as applied to listed rms
categorized into three sectors namely the manufacturing sector, the retail sector and
the nancial sector rms. Here again, the dominance of the US dollar is evident. Two
out of the three sectors (manufacturing and retail) have signicant exposures to the US
dollar exchange risk. The direction of exposure however differs. Whilst the retail sector
is negatively exposed to the US dollar exchange risk, the manufacturing sector is
positively exposed to the US dollar exchange risk. The results of the retail sector
exposure to the changes in the US dollar are not surprising since rms in the retail
sector mainly deal in imported goods. Therefore we expect the depreciation of the value
of the cedi relative to the US dollar to make imports relatively expensive as more cedis
are required to import the same units of items.
JRF
8,4
386
F
i
r
m
C
o
n
s
t
a
n
t
U
S
d
o
l
l
a
r
E
u
r
o
U
K
p
o
u
n
d
R
m
R
2
F
-
s
t
a
t
.
f
-
p
r
o
b
.
N
P
r
o
b
a
b
i
l
i
t
y
v
a
l
u
e
s
a
N
P
r
o
b
a
b
i
l
i
t
y
v
a
l
u
e
s
a
N
P
r
o
b
a
b
i
l
i
t
y
v
a
l
u
e
s
a
N
P
r
o
b
a
b
i
l
i
t
y
v
a
l
u
e
s
a
N
P
r
o
b
a
b
i
l
i
t
y
v
a
l
u
e
s
a
N
N
N
H
F
C
2
0
.
2
2
9
0
.
8
6
5
0
2
6
.
9
1
2
3
0
.
0
9
9
*
1
3
.
5
8
4
0
.
7
2
9
1
1
4
.
7
5
2
0
.
0
5
2
*
*
1
.
3
1
0
4
0
.
0
0
*
*
*
0
.
4
3
9
1
2
.
9
2
0
.
0
0
0
E
I
C
5
.
1
4
3
2
0
.
0
3
8
2
3
1
.
1
4
2
0
.
0
4
2
*
*
5
1
.
6
6
9
0
.
0
6
8
*
*
2
2
7
.
7
9
8
0
.
7
9
2
0
.
5
3
1
0
.
1
3
6
0
.
5
4
9
1
1
.
8
5
0
.
0
0
8
F
M
L
4
.
3
6
4
0
.
0
1
4
2
9
.
7
6
0
4
2
1
0
.
0
5
9
*
7
.
5
1
9
0
.
0
2
4
*
*
1
.
0
5
1
0
.
9
8
8
0
.
5
9
7
0
.
0
2
6
*
*
0
.
4
2
5
1
2
.
4
1
0
.
0
2
1
M
G
L
0
.
0
0
1
0
.
9
7
4
2
5
.
4
1
5
0
.
0
2
7
*
*
7
.
7
5
8
0
.
3
5
2
2
1
.
9
4
7
0
.
1
3
4
0
.
7
9
0
0
.
0
0
7
*
*
*
0
.
4
8
0
1
5
.
4
6
0
.
0
0
9
M
O
G
L
0
.
7
5
3
0
.
3
7
2
2
8
.
2
2
9
0
.
0
1
1
*
*
*
9
.
4
4
9
0
.
7
9
5
1
0
.
0
2
7
0
.
0
2
3
*
*
0
.
4
1
9
0
.
0
0
1
*
*
*
0
.
4
6
2
1
4
.
3
7
0
.
0
1
9
P
A
F
2
.
0
1
7
0
.
1
9
1
2
2
.
1
9
2
0
.
0
1
4
*
*
*
2
4
.
5
3
2
0
.
3
2
1
5
.
1
5
5
0
.
0
0
8
*
*
*
0
.
4
2
4
4
7
0
.
0
1
2
*
*
*
0
.
5
2
0
1
8
.
6
9
0
.
0
0
1
S
C
B
2
.
2
3
7
0
.
1
7
5
3
2
.
7
3
9
0
.
6
0
6
2
0
.
3
9
5
0
.
6
6
9
2
6
2
.
8
4
6
0
.
3
7
6
0
.
9
1
8
0
.
0
0
0
*
*
*
0
.
2
0
6
4
.
2
9
0
.
0
0
4
S
P
P
C
1
.
8
4
7
0
.
2
1
1
2
3
.
1
5
4
0
.
0
3
1
*
*
2
2
.
6
4
2
0
.
5
8
2
4
.
9
2
5
0
.
0
4
5
*
*
0
.
1
2
7
0
.
0
1
1
*
*
*
0
.
3
2
0
7
.
9
0
0
.
0
0
6
S
S
B
2
.
3
5
7
0
.
1
6
0
2
1
0
.
0
7
2
0
.
0
1
6
*
*
2
2
.
6
4
1
0
.
0
5
6
*
5
.
0
1
7
0
.
0
2
5
*
*
1
.
1
2
1
0
.
0
0
0
*
*
*
0
.
2
5
8
5
.
7
3
0
.
0
0
0
U
N
I
L
2
0
.
0
1
9
0
.
9
9
9
2
8
.
3
2
5
0
.
0
8
8
*
2
.
5
0
9
0
.
7
7
3
8
.
1
2
9
0
.
0
8
0
*
5
.
3
1
9
0
.
0
1
3
*
*
0
.
5
6
4
2
1
.
7
2
0
.
0
0
9
B
A
T
6
.
8
3
7
0
.
0
0
0
0
.
2
7
7
0
.
6
9
3
0
.
2
0
7
0
.
7
4
3
2
1
.
1
2
1
0
.
1
8
5
0
.
5
7
9
0
.
0
2
8
*
*
0
.
5
1
6
1
7
.
8
4
0
.
0
0
8
C
F
A
O
2
.
8
9
2
0
.
0
0
0
2
0
.
8
2
4
0
.
6
9
3
2
0
.
0
5
1
0
.
7
4
3
0
.
5
1
3
0
.
1
8
5
0
.
0
5
9
0
.
0
2
8
*
*
0
.
6
2
3
2
8
.
7
1
0
.
0
0
8
A
L
W
0
.
1
9
1
0
.
9
5
3
1
.
2
8
8
0
.
3
2
1
2
1
.
5
3
9
0
.
1
9
2
0
.
4
6
2
0
.
7
6
6
1
.
5
5
1
0
.
0
0
2
*
*
*
0
.
1
5
4
2
.
9
9
0
.
0
2
4
G
G
L
2
0
.
7
9
8
0
.
2
0
3
3
.
5
6
4
0
.
0
2
4
*
*
0
.
7
1
9
0
.
2
0
6
2
0
.
3
9
8
0
.
5
9
4
0
.
8
4
5
0
.
0
0
0
*
*
*
0
.
4
4
1
1
3
.
2
6
0
.
0
0
0
G
C
B
0
.
1
3
1
0
.
4
0
1
2
0
.
1
0
8
6
0
.
0
8
0
*
2
0
.
0
0
9
0
.
8
6
4
3
0
.
1
2
2
0
.
0
9
5
*
0
.
0
4
6
6
1
0
.
0
1
8
*
*
0
.
4
5
9
0
.
9
9
0
.
4
1
7
P
Z
3
.
4
4
7
0
.
0
1
8
2
0
.
7
0
1
0
.
2
2
0
1
.
0
9
5
0
.
0
3
7
*
*
2
0
.
8
6
9
0
.
2
0
4
0
.
0
4
9
0
.
8
1
6
0
.
1
0
2
1
.
8
6
0
.
1
2
6
M
L
C
0
.
9
2
3
0
.
7
0
0
0
.
5
3
3
0
.
5
7
9
2
0
.
1
8
2
0
.
8
3
5
2
0
.
7
8
8
0
.
4
9
4
2
.
0
6
3
0
.
0
0
0
*
*
*
0
.
3
5
4
9
.
0
5
0
.
0
0
0
C
P
C
1
.
7
8
2
0
.
1
9
5
0
.
0
7
5
0
.
8
9
0
2
0
.
9
3
5
0
.
0
6
3
*
0
.
4
5
9
0
.
4
8
5
0
.
0
2
4
0
.
9
0
5
0
.
0
6
8
1
.
1
4
0
.
3
4
7
C
M
L
T
0
.
9
2
4
0
.
0
9
7
2
0
.
3
5
5
0
.
1
1
2
0
.
2
2
4
0
.
2
6
6
0
.
0
3
4
0
.
8
9
7
0
.
2
9
3
0
.
0
0
0
*
*
*
0
.
2
3
0
4
.
9
4
0
.
0
0
1
A
B
L
2
2
.
2
3
5
0
.
1
0
2
0
.
8
2
9
0
.
1
2
9
0
.
1
2
8
0
.
7
9
4
2
0
.
3
9
3
0
.
5
4
5
0
.
4
7
1
0
.
0
2
1
*
*
0
.
1
1
2
2
.
0
9
0
.
0
9
2
N
o
t
e
s
:
a
P
r
o
b
a
b
i
l
i
t
y
v
a
l
u
e
s
o
f
t
h
e
t
-
s
t
a
t
i
s
t
i
c
s
;
*
,
*
*
,
a
n
d
*
*
*
i
n
d
i
c
a
t
e
s
s
t
a
t
i
s
t
i
c
a
l
s
i
g
n
i

c
a
n
c
e
a
t
t
h
e
1
0
%
,
5
%
a
n
d
1
%
r
e
s
p
e
c
t
i
v
e
l
y
.
R
e
g
r
e
s
s
i
o
n
a
n
a
l
y
s
i
s
(
O
L
S
)
f
o
r
e
x
c
h
a
n
g
e
-
r
a
t
e
e
x
p
o
s
u
r
e
c
o
e
f

c
i
e
n
t
s
.
T
h
e
d
a
t
a
r
e
p
r
e
s
e
n
t
s
m
o
n
t
h
l
y
o
b
s
e
r
v
a
t
i
o
n
s
o
f
2
0

r
m
s
c
o
r
r
e
c
t
e
d
f
o
r
a
u
t
o
c
o
r
r
e
l
a
t
i
o
n
a
n
d
h
e
t
e
r
o
s
c
e
d
a
s
t
i
c
i
t
y
,
f
r
o
m
1
9
9
9
t
o
2
0
0
4
Table I.
Firm exposure to multiple
exchange rates
Foreign
exchange risk
exposure
387
The positive exposure of the US dollar in the case of the manufacturing sector means
that the manufacturing rms rather gain from the depreciation of the exchange rate.
This means that the depreciation of the cedi in relation to US dollar exchange rate
makes imported goods more expensive and this may result in a shift in demand to local
substitutes.
Finally sector specic results based on the trade weighted exchange rate index are
presented in Table IV. Here only the retail sector is signicantly exposed to the NEER.
The sector has a positive exposure coefcient to the NEER, implying that the sector
gains from a depreciation of the NEER. Though this is a deviation from the earlier
results for the retail sector, the positive exposure could be due to the combined effects
of the different weights in the NEER. Changes in the trade weighted index do not
appear to be signicant in explaining the performance of rms in the manufacturing
and nancial sectors.
When we compare the sector specic exposure results to that of the individual
currencies, we nd that, although similar rms in the same industry may individually
have a signicant exchange rate coefcient, aggregating the rms into sectors tend to
affect the magnitude and direction of exposure. This is possible because two rms in
the same industry for example could be strongly affected by exchange rate movements,
but one rm may be beneting in terms of higher returns from an appreciation in the
value of the domestic currency whilst the returns of another rm is reduced by such an
appreciation.
Firm NEER F-stat. f-prob.
N Probability values of t-statistics N N
HFC 0.066 0.057
* *
13.92 0.000
EIC 20.0911 0.083
*
21.85 0.000
FML 0.009 0.853 16.51 0.071
MGL 20.000 0.954 18.66 0.009
MOGL 20.014 0.536 26.37 0.029
PAF 20.004 0.914 38.69 0.001
SCB 20.085 0.036
* *
6.29 0.004
SPPC 20.052 0.052
* *
8.90 0.006
SG-SSB 20.211 0.615 9.73 0.000
UNIL 20.431 0.485 21.72 0.009
BAT 20.201 0.661 19.84 0.008
CFAO 0.010 0.698 38.71 0.008
ALW 0.449 0.0000
* * *
2.99 0.024
GGL 20.007 0.864 1.83 0.419
GCB 20.000 0.842 19.26 0.000
PZ 20.003 0.923 2.86 0.126
MLC 0.263 0.000
* * *
19.05 0.000
CPC 20.004 0.908 11.16 0.000
CMLT 0.024 0.097
*
10.94 0.001
ABL 0.012 0.744 1.09 0.522
Notes:
*
,
* *
, and
* * *
Indicates statistical signicance at the 10%, 5%, and 1% respectively.
Regression analysis (OLS) for exchange-rate exposure coefcients. The data represents monthly
observations of 20 rms corrected for autocorrelation and heteroscedasticity, from 1999 to 2004
Table II.
Exposure to the trade
weighted exchange rate
(NEER)
JRF
8,4
388
C
o
n
s
t
a
n
t
U
S
d
o
l
l
a
r
E
u
r
o
U
K
p
o
u
n
d
R
m
S
e
c
t
o
r
N
P
r
o
b
a
b
i
l
i
t
y
v
a
l
u
e
s
o
f
t
-
s
t
a
t
i
s
t
i
c
s
N
P
r
o
b
a
b
i
l
i
t
y
v
a
l
u
e
s
o
f
t
-
s
t
a
t
i
s
t
i
c
s
N
P
r
o
b
a
b
i
l
i
t
y
v
a
l
u
e
s
o
f
t
-
s
t
a
t
i
s
t
i
c
s
N
P
r
o
b
a
b
i
l
i
t
y
v
a
l
u
e
s
o
f
t
-
s
t
a
t
i
s
t
i
c
s
N
P
r
o
b
a
b
i
l
i
t
y
v
a
l
u
e
s
o
f
t
-
s
t
a
t
i
s
t
i
c
s
R
e
t
a
i
l
1
.
7
9
3
0
.
0
6
0
2
0
.
6
4
2
0
.
0
3
9
*
*
2
0
.
0
6
1
0
.
8
6
2
2
0
.
9
7
9
0
.
0
8
6
*
0
.
8
1
1
0
.
0
0
0
*
*
*
M
a
n
u
f
a
c
t
u
r
i
n
g
3
.
0
7
3
0
.
0
6
0
5
.
8
1
5
0
.
0
3
0
*
*
2
6
.
1
9
8
0
.
1
0
2
0
.
1
0
2
0
.
7
9
8
1
.
6
1
2
0
.
5
0
8
F
i
n
a
n
c
i
a
l
1
3
.
4
9
7
0
.
0
0
1
2
0
.
6
5
8
0
.
0
5
0
*
2
0
.
2
0
4
2
0
.
5
1
2
0
.
5
4
7
0
.
1
8
5
0
.
6
7
5
0
.
0
0
0
*
*
*
N
o
t
e
s
:
*
,
*
*
,
a
n
d
*
*
*
I
n
d
i
c
a
t
e
s
s
t
a
t
i
s
t
i
c
a
l
s
i
g
n
i

c
a
n
c
e
a
t
t
h
e
1
0
%
,
5
%
,
a
n
d
1
%
r
e
s
p
e
c
t
i
v
e
l
y
.
R
e
g
r
e
s
s
i
o
n
a
n
a
l
y
s
i
s
(
O
L
S
)
f
o
r
e
x
c
h
a
n
g
e
-
r
a
t
e
e
x
p
o
s
u
r
e
c
o
e
f

c
i
e
n
t
s
.
T
h
e
d
a
t
a
r
e
p
r
e
s
e
n
t
s
m
o
n
t
h
l
y
o
b
s
e
r
v
a
t
i
o
n
s
o
f
2
0

r
m
s
c
o
r
r
e
c
t
e
d
f
o
r
a
u
t
o
c
o
r
r
e
l
a
t
i
o
n
a
n
d
h
e
t
e
r
o
s
c
e
d
a
s
t
i
c
i
t
y
,
f
r
o
m
1
9
9
9
t
o
2
0
0
4
Table III.
Sector specic exposures
to multiple exchange
rates
Foreign
exchange risk
exposure
389
Conclusions
The study examined the exposure of listed rms on the Ghana Stock Exchange to
exchange rate risk for the period January 1999 to December 2004. The ndings of the
study are that, all the major currencies of international transaction of the country are
sources of foreign exchange risk to listed rms on the GSE. The US dollar turned out to
be the most dominant source of exchange rate risk at both the rm and sector levels.
Most rms had negative exposure coefcients and this suggests that, the majority of
the listed rms could experience an adverse valuation effect when the local currency
(cedi) depreciates substantially against other foreign currencies and benet when the
cedi strengthens in value relative to these currencies.
Firms with exposure to the UK pound on the other hand experienced higher stock
returns as the cedi depreciated against the UK pound. It is therefore advantageous for
listed rms on the GSE to denominate their international transactions with the UK
pound.
At the sector level, there is evidence of risk exposure particularly in the US dollar
with a positive exposure in the manufacturing sector and negative exposure in the
retail sector. In general, most listed rms on the Ghana Stock Exchange are
signicantly exposed to foreign exchange risk emanating from all the three major
currencies of international trade, namely, the US dollar, the UK pound and the euro.
The practical relevance of our ndings in foreign exchange management lies in the
fact that, even though there are a number of techniques such as balance sheet hedging,
use of derivatives, leading and lagging amongst others available to manage foreign
exchange risk in most developed countries, these measures tend to be rather too
sophisticated and difcult to implement in developing countries like Ghana with
undeveloped nancial systems. Nonetheless, given the degree of exposure revealed in
this study, corporate managers and investors in Ghana should endeavour to apply a
combination of simple tools such as the use of forward contracts and swaps to
supplement price adjustments and investment in foreign currency in order to minimize
their exposure to exchange risk. Despite the short-comings of the nancial system in
terms of availability of tools for managing foreign exchange risk exposure,
instruments are still available to manage the risk exposure. Abor (2005) for instance
nds that most rms in Ghana manage their foreign exchange risk through two means;
the use of price adjustments to reect balance sheet changes and the buying and
holding of foreign currency.
Finally, although the Ghana government and the central bank authorities are
continuously applying prudent macroeconomic policies to ensure stability in the
NEER Rm
Sector N Probability values of t-statistics N Probability values of t-statistics
Retail 0.086 0.000
* * *
0.758 0.000
* * *
Manufacturing 20.112 0.794 1.348 0.598
Financial 0.008 0.564 0.699 0.000
* * *
Notes:
*
,
* *
, and
* * *
Indicates statistical signicance at the 10%, 5%, and 1% respectively.
Regression analysis (OLS) for exchange-rate exposure coefcients. The data represents monthly
observations of 20 rms corrected for autocorrelation and heteroscedasticity, from 1999 to 2004
Table IV.
Sector specic exposures
to the trade-weighted
exchange rate (NEER)
JRF
8,4
390
exchange rate, the exchange rate risk faced by rms forms a signicant component of
their risk prole. It is therefore imperative that listed rms and generally all rms in
Ghana with international operations effectively manage their risk to minimize their
exposure to exchange rate risk.
Future studies could investigate the effect of the use of different foreign exposure
minimization mechanisms on rm valuation and stock returns. It would also be
interesting for future studies to examine the importance of rm size and location in
foreign exchange exposure management and rm performance.
Notes
1. The domestic currency of Ghana.
2. Dividend income is ignored in the computation of the monthly rm returns since in Ghana
dividends are usually paid at the end of the year. Therefore returns for each rm is computed
based on the changes in share prices- capital gains or losses.
3. The cedi to euro is computed for the periods 2002 -2004 since the euro only became a legal
tender in 2002.
4. We do not lay much emphasis on the results of the coefcients of the cedi to euro due to the
loss in data between 1999 and 2002 when the euro was not legal tender.
References
Aryeetey, E. and Tarp, F. (2000), Structural adjustment and after: which way forward?, in
Aryeetey, E., Harrigan, J. and Nissanke, M. (Eds), Economic Reforms in Ghana: The
Miracle and the Mirage, James Currey, Woeli Publishing, Africa World Press, Oxford,
Accra, Trenton, NJ.
Abor, J. (2005), Managing foreign exchange risk among Ghanaian rms, Journal of Risk
Finance, Vol. 6 No. 4, pp. 306-18.
Adler, M. and Dumas, B. (1984), Exposure to currency risk: denition and measurement,
Financial Management, Summer, pp. 40-50.
Amin, A.A. (1996), The effects of exchange rate policy on Cameroons agricultural
competitiveness, Research Paper 42, AERC.
Atta, J.K., Jefferis, K.R., Mannathoko, I. and Siwawa, N. (1999), Exchange rate policy and price
determination in Botswana, Research Paper 93, AERC.
Bartov, E. and Bodnar, G.M. (1994), Firm valuation, earnings expectations, and the
exchange-rate exposure effect, Journal of Finance, Vol. 49 No. 5, pp. 1755-85.
Bodnar, G.M. and Gentry, W.M. (1993), Exchange rate exposure and industry characteristics:
evidence from Canada, Japan, and the USA, Journal of International Money and Finance,
Vol. 12, February, pp. 29-45.
Bodnar, G.M. and Wong, M.H.F. (2000), Estimating exchange rate exposure: some weighty
issues, Working Paper 7497, NBER, January.
Bodnar, G.M. and Wong, M.H.F. (2003), Estimating exchange rate exposures: issues in model
structure, Financial Management, Spring, pp. 35-67.
Booth, L. and Rotenberg, W. (1990), Assessing foreign exchange exposure: theory and
application using Canadian rms, Journal of International Financial Management and
Accounting, Vol. 2, June, pp. 1-22.
Chamberlain, S., Howe, J. and Popper, H. (1997), The exchange rate exposure of US and Japanese
banking institutions, Journal of Banking and Finance, Vol. 21, pp. 871-92.
Foreign
exchange risk
exposure
391
Choi, J.J. and Prasad, A.M. (1995), Exchange risk sensitivity and its determinants: a rm and
industry analysis of US multinationals, Financial Management, Vol. 24 No. 3, pp. 77-88.
Choi, J.J. and Elyasiani, E. (1997), Derivative exposure and the interest rate and exchange rate
risks of US banks, Journal of Financial Services Research, Vol. 12, pp. 267-86.
Choi, J.J., Elyasiani, E. and Kopecky, K.J. (1992), The sensitivity of bank stock returns to market,
interest and exchange rate risks, Journal of Banking and Finance, Vol. 16, pp. 983-1004.
Chow, E.H., Lee, W. and Solt, M. (1997a), The exchange-rate risk exposure of asset returns,
Journal of Business, Vol. 70, pp. 105-23.
Chow, E.H., Lee, W. and Solt, M. (1997b), The economic exposure of us multinational rms,
Journal of Financial Research, Vol. 2, pp. 191-210.
Dominguez, K. and Tesar, L. (2001), A re-examination of exchange rate exposure, American
Economic Review Papers and Proceedings, Vol. 91.
Dordunoo, C.K. (1994), The foreign exchange market and the Dutch auction system in Ghana,
Research Paper 24, AERC.
Fang, H. and Loo, J.C.H. (1994), Dollar value and stock returns, International Review of
Economics and Finance, Vol. 3 No. 2, pp. 221-31.
Glaum, M. (1990), Strategic management of exchange rate risks, Long Range Planning, Vol. 23
No. 4, pp. 65-72.
Grant, R. and Soenen, L. (1991), Conventional hedging: an inadequate response to long-term
foreign exchange exposure, Managerial Finance, Vol. 17 No. 4, pp. 1-4.
He, J. and Ng, L.K. (1998), The foreign exchange exposure of Japanese multinational,
Corporations Journal of Finance, Vol. 53, pp. 733-53.
Heckman, C.R. (1983), Measuring foreign exchange exposure: a practical theory and its
application, Financial Analysts Journal, Vol. 39, pp. 59-65.
Jorion, P. (1990), The exchange-rate exposure of US multinationals, Journal of Business, Vol. 63
No. 3, pp. 331-45.
Jorion, P. (1991), The pricing of exchange rate risk in the stock market, Journal of Financial and
Quantitative Analysis, Vol. 26 No. 3, pp. 363-76.
Khoo, A. (1994), Estimation of foreign exchange exposure: an application to mining companies
in Australia, Journal of International Money and Finance, Vol. 13 No. 3, pp. 342-63.
Loudon, G. (1993), The foreign exchange operating exposure of Australian stocks, Accounting
and Finance, Vol. 33 No. 1, pp. 19-32.
Madura, J. (2003), International Financial Management, 7th ed., South-Western College
Publishing, Boston, MA.
Miller, K. and Reuer, J. (1998), Firm strategy and economic exposure to foreign exchange rate
movements, Journal of International Business Studies, Vol. 29, third quarter, pp. 493-514.
Ndunguu, N. (1997), Price and exchange rate dynamics in Kenya: an empirical investigation
(1970-1993), Research Paper 58, AERC.
Ndunguu, N. (1999), Monetary and exchange rate policy in Kenya, Research Paper 94, AERC.
Odubogun, K. (1995), Institutional reforms and the management of exchange rate policy in
Nigeria, Research Paper 36, AERC.
Oduro, A.D. and Harrigan, J. (2000), Exchange rate policy and the balance of payments, in
Aryeetey, E., Harrigan, J. and Nissanke, M. (Eds), Economic Reforms in Ghana; The
Miracle and the Mirage, James Currey, Woeli Publishing, Africa World Press,
Oxford/Accra/Trenton, NJ.
JRF
8,4
392
Ogun, O. (1998), Real exchange rate movements and export growth: Nigeria, 1960-1990,
Research Paper 82, AERC.
Parsley, D. and Popper, H. (2002), Exchange rate pegs and foreign exchange exposure,
Working Paper No. 0218, Vanderbilt University, Financial Markets Research Center.
Rutasitara, L. (2004), Exchange rate regimes and ination in Tanzania, Research Paper 138,
AERC.
Silumbu, E.B.D. (1995), The role of exchange rate and monetary policy in the monetary
approach to the balance of payments: evidence from Malawi, Research Paper 37, AERC.
Corresponding author
Zubeiru Salifu can be contacted at: zubbess@yahoo.co.uk
Foreign
exchange risk
exposure
393
To purchase reprints of this article please e-mail: reprints@emeraldinsight.com
Or visit our web site for further details: www.emeraldinsight.com/reprints
Banks risk management:
a comparison study of UAE
national and foreign banks
Hussein A. Hassan Al-Tamimi
Department of Business Administration, College of Business Administration,
University of Sharjah, United Arab Emirates, and
Faris Mohammed Al-Mazrooei
Abu-Dhabi, United Arab Emirates
Abstract
Purpose The purpose of this research is to examine the degree to which the UAE banks use risk
management practices and techniques in dealing with different types of risk. The secondary objective
is to compare risk management practices between the two sets of banks.
Design/methodology/approach The authors developed a modied questionnaire, divided into
two parts. The rst part covers six aspects: understanding risk and risk management; risk
identication; risk assessment and analysis; risk monitoring; risk management practices; and credit
risk analysis. This part includes 43 closed-ended questions based on an interval scale. The second part
consists of two closed-ended questions based on an ordinal scale dealing with two topics: methods of
risk identication, and risks facing the sample banks.
Findings This study found that the three most important types of risk facing the UAE commercial
banks are foreign exchange risk, followed by credit risk, then operating risk. It also found that the
UAE banks are somewhat efcient in managing risk, and risk identication and risk assessment and
analysis are the most inuencing variables in risk management practices. Finally, the results indicate
that there is a signicant difference between the UAE national and foreign banks in the practice of risk
assessment and analysis, and in risk monitoring and controlling.
Originality/value The article will be of value to those interested in the banking industry.
Keywords Risk management, Risk analysis, Credit management, United Arab Emirates, Banks,
Foreign exchange
Paper type Research paper
Introduction
Risk management is a cornerstone of prudent banking practice. Undoubtedly all banks
in the present-day volatile environment are facing a large number of risks such as
credit risk, liquidity risk, foreign exchange risk, market risk and interest rate risk,
among others risks which may threaten a banks survival and success. In other
words, banking is a business of risk. For this reason, efcient risk management is
absolutely required. Carey (2001) indicates in this regard that risk management is more
important in the nancial sector than in other parts of the economy. The purpose of
nancial institutions is to maximize revenues and offer the most value to shareholders
by offering a variety of nancial services, and especially by administering risks.
Recently many commercial banks have appointed senior managers to oversee a formal
risk management function.
Risk can be classied into systematic and unsystematic risk. Systematic risk is
associated with the overall market or the economy, whereas unsystematic risk is
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1526-5943.htm
JRF
8,4
394
The Journal of Risk Finance
Vol. 8 No. 4, 2007
pp. 394-409
qEmerald Group Publishing Limited
1526-5943
DOI 10.1108/15265940710777333
related to a specic asset or rm. Some of the systematic risk can be reduced through
the use of risk mitigation and transmission techniques. In this regard Oldeld and
Santomero (1997) refer to three generic risk-mitigation strategies:
(1) eliminate or avoid risks by simple business practices;
(2) transfer risks to other participants; and
(3) actively manage risks at the bank level (acceptance of risk).
Thus, nancial intermediaries may avoid specic risks by simplifying business
practices and minimizing activities that inict risk. Activities with which the nancial
institution is committed to proceed can be adeptly managed or transferred. Certain
risks which are inevitable or transferred must be engulfed by the bank. Inevitable risks
are those too complex to separate from assets. The subsequent risk is accepted by the
bank as being crucial to its business; banks are specialized in dealing with this sort of
risk, and reap the benets. According to the consultative paper issued by the Basel
Committee on Banking Supervision (1999), for most banks loans are the largest and
most obvious source of credit risk. Banks are increasingly facing credit in various
nancial instruments other than loans, including acceptances, interbank transactions,
trade nancing, foreign exchange transactions, nancial futures, swaps, bonds,
equities, options, the extension of commitments and guarantees, and the settlement of
transactions. The Basel II regulation is likely to add fuel in this regard because new
rules on how much capital banks must hold, will make a few lending decisions even
more protable.
The objective of this study is to examine the degree to which the UAE banks use risk
management practices and techniques in dealing with different types of risk, and effective
risk management practices followed by the UAE national and foreign banks. The
objective also is to compare risk management practices between the two sets of banks.
To the best of the authors knowledge most UAE banks suffer from loan default
problems. In addition to credit risk, these banks face other types of risk: foreign exchange
risk, market risk and interest rate risk, among others. The current study also intends to
identify the most important type or types of risk facing the UAE commercial banks
Literature review
There have been a large number of studies published about risk management in
general. However, the number of the empirical studies on risk management practices in
nancial institutions was found to be relatively small. The following is an attempt to
summarize the main conclusions of some selected studies.
Linbo Fan (2004) examined efciency versus risk in large domestic USA banks. He
found that prot efciency is sensitive to credit risk and insolvency risk but not to
liquidity risk or to the mix of loan products. Ho Hahm (2004) conducted an empirical
study on interest rate and exchange rate exposures of banking institutions in pre-crisis
Korea. Results indicated that Korean commercial banks and merchant banking
corporations had been signicantly exposed to both interest rate and exchange rate
risks, and that the subsequent protability of commercial banks was signicantly
associated with the degree of pre-crisis exposure. The results also indicated that the
Korean case highlights the importance of upgrading nancial supervision and risk
management practices as a precondition for successful nancial liberalization.
Niinimaki (2004) in his paper entitled The effects of competition on banks risk
taking found that the magnitude of risk taking depends on the structure and side of
Banks risk
management
395
the market in which competition takes place. He also concluded that if the bank is a
monopoly or banks are competing only in the loan market, deposit insurance has no
effect on risk taking. Banks in this situation tend to take risks, although extreme risk
taking is avoided. In contrast, introducing deposit insurance increases risk taking if
banks are competing for deposits. In this case, deposit rates become excessively high,
thereby forcing banks to take extreme risks. Wetmore (2004) examined the relationship
between liquidity risk and loans-to-core deposits ratio of large commercial bank
holding companies. He concluded that the average loan-to-core deposit ratio had
increased over the period studied, which reects a change in the asset/liability
management practices of banks. He also concluded that there is a positive relationship
occurring between market risk and the change in loan-to-core deposits ratio after 1994,
with a negative relationship occurring before 1994.
Wang and Sheng-Yung (2004) studied foreign exchange risk, world diversication and
Taiwanese American depository receipts (ADRs). In this study they tried to answer the
following question: Should USA investors purchase American depository receipts issued
by Taiwanese multinationals? Empirical results indicated that foreign exchange risk is
priced in Taiwanese ADRs. Moreover, Taiwanese ADRs were shown to help USA
investors diversify their portfolios globally. These ndings suggest that Taiwanese ADRs
are valid investment tools for USA investors who seek international diversications.
Khambata and Bagdi (2003) examined off-balance-sheet (OBS) credit risk across the
top 20 Japanese banks. The main results of this study indicated that nancial
derivatives are heavily used by the top four banks and that loan commitments are the
largest source of credit risk among traditional OBS instruments. The results also
indicated that there is a wide difference across the banks in the use of derivative
leverage. As compared to USA and European banks, Japanese banks use fewer OBS
instruments as a percentage of their assets. This implies that Japanese banks are more
conservative and risk-averse in general than their USA or European counterparts,
especially given the bad nancial condition of Japanese banks.
Al-Tamimi (2002) investigated the degree to which the UAE commercial banks use
risks management techniques in dealing with different types of risk. The study found
that the UAE commercial banks were mainly facing credit risk. The study also found
that inspection by branch managers and nancial statement analysis were the main
methods used in risk identication. The main techniques used in risk management
according to this study were establishing standards, credit score, credit worthiness
analysis, risk rating and collateral; the study also highlighted the willingness of the
UAE commercial banks to use the most sophisticated risk management techniques,
and recommended the adoption of a conservative credit policy.
Salas and Saurina (2002) examined credit risk in Spanish commercial and savings
banks; they used panel data to compare the determinants of problem loans of Spanish
commercial and savings banks in the period 1985-1997, taking into account both
macroeconomic and individual bank-level variables. The GDP growth rate, rms,
family indebtedness, rapid past credit or branch expansion, inefciency, portfolio
composition, size, net interest margin, capital ratio and market power are variables that
explain credit risk. Their ndings raise important bank supervisory policy issues: the
use of bank-level variables as early warning indicators, the advantages of mergers of
banks from different regions, and the role of banking competition and ownership in
determining credit risk.
Oldeld and Santomero (1997) investigated risk management in nancial institutions.
In this study, they suggested four steps for active risk management techniques:
JRF
8,4
396
(1) the establishment of standards and reports;
(2) the imposition of position limits and rules (i.e. contemporary exposures, credit
limits and position concentration);
(3) the creation of self investment guidelines and strategies; and
(4) the alignment of incentive contracts and compensation (performance-based
compensation contracts).
Based on the literature review above, the following comments can be made:
.
Prot efciency is sensitive to credit risk and insolvency risk but not to liquidity
risk or to the mix of loan products.
.
Risk management practices are a precondition for successful nancial
liberalization.
.
Commercial banks are mainly faced with credit risk; loans are the largest and
most obvious source of this type of risk.
.
Inspection by branch managers and nancial statement analysis are the main
methods used in risk identication.
.
The appropriate risk weight for an off-balance-sheet contract is likely to depend
on the size of the bank.
Research methodology
Research questions and hypotheses
This study attempts to answer the following questions:
RQ1. Do the UAE banks staff understand risk and risk management?
RQ2. Have the UAE banks clearly identied the potential risks relating to each of
their declared aims and objectives?
RQ3. Do the UAE banks efciently assess and analyze risk in general?
RQ3. Do the UAE banks have an efcient risk monitoring and controlling system?
RQ3. Do the UAE banks have efcient risk management?
RQ3. What are the methods used in risk management in general?
Based on the stated purpose and the questions mentioned above, the following
hypotheses are formulated:
H1. There is a positive relationship between risk management practices and
understanding risk, and risk management, risk identication, risk assessment
and analysis, risk monitoring, risk, and credit risk analysis.
H2. There is a difference between the UAE national and foreign banks in the
understanding of risk and risk management.
H3. There is a difference between the UAE national and foreign banks in the
practices of risk identication.
H4. There is a difference between the UAE national and foreign banks in the
practices of risk assessment and analysis.
Banks risk
management
397
H5. There is a difference between the UAE national and foreign banks in risk
monitoring and controlling.
H6. There is a difference between the UAE national and foreign banks in the
context of risk management practices.
H7. There is a difference between the UAE national and foreign banks in the
practice of credit risk analysis.
Instrument
The authors developed a modied questionnaire, divided into two parts. The rst part
covers six aspects: understanding risk and risk management; risk identication; risk
assessment and analysis; risk monitoring; risk management practices; and credit risk
analysis. This part includes 43 closed-ended questions based on an interval scale,
where eight questions correspond to the understanding risk and risk management
aspect, ve questions correspond to risk identication, seven questions correspond to
risk assessment and analysis, six questions correspond to risk monitoring, ten
questions correspond to risk management practices, and seven questions correspond to
credit risk analysis. Respondents were asked to indicate their degree of agreement with
each of the questions on a seven-point Likert scale.
The second part consists of two closed-ended questions based on an ordinal scale
dealing with two topics: methods of risk identication, and risks facing the sample
banks. The responses were classied into national and foreign banks. The aim of this
classication is to nd out if there are any differences between the national banks and
foreign banks regarding risk management practices. It is worth mentioning here that
most foreign banks represent branches of highly sophisticated banks, and so it is
assumed that these banks use sophisticated risk management techniques.
To assess the scales content validity the author asked six experts three
academicians and three practitioners to examine the scales, as was suggested by
Devellis (1991). Accordingly, the author made some changes to the rst draft in terms
of eliminating, adding to or rewording some of the questions included in that draft.
Sampling and data collection
The UAE has 46 commercial banks, 21 of which are national banks and the remaining
25 of which are foreign banks (branches of banks located outside the UAE). The
targeted population are those heavily involved in risk management among the major
conventional banks whose total assets exceed AED 7 billion or US$ 1.9 billion in the
three largest cities of the UAE, namely, Abu Dhabi, Dubai and Sharjah. These cities
host more than 80 per cent of the UAE population. The sample includes eight national
conventional banks: Abu Dhabi National Bank, Dubai National Bank, Abu Dhabi
Commercial Bank, Mashriq Bank, Dubai International Bank, the National Union Bank,
First Gulf Bank and Commercial Bank of Dubai. The sample also includes the ve
largest foreign banks: HSBC Bank Middle East, Bank Saderat Iran, Citi Bank, Standard
Chartered Bank and Arab Bank.
Regarding the four Islamic banks (Dubai Islamic Bank, Abu Dhabi Islamic,
Emirates Islamic Bank, and Sharjah Islamic Bank), the authors decided to distribute
questionnaires to all ofcers involved in risk management in the three cities because of
the small number of the Islamic banks. The total assets of the four Islamic banks was
JRF
8,4
398
about 48,904 billion (US$ 12.485 billion), constituting around 11.1 per cent of the total
assets of the UAE commercial banks in 2004 (Emirates Banks Association, 2004).
The authors handed questionnaires to the banks head ofces and branch managers
who were requested to kindly pass the questionnaires to the targeted ofcers. The
questionnaires were distributed to those ofcers involved in risk management
activities, namely, branch managers, senior risk management ofcers and senior credit
ofcers. From the 188 questionnaires we received, the screening process resulted in
excluding 31 responses from the study because of missing data questions. The
remaining 157 responses represent an effective response rate of around 49 per cent of
the total sample (the number of ofcers involved in risk management is around 319). Of
these respondents, 73.9 per cent were ofcers from national banks and 26.1 per cent
were ofcers from foreign banks. The high percentage of respondents from national
banks can be explained by the fact that the total assets of national banks constituted
about 76.5 per cent of the assets of UAE commercial banks in 2004.
Data analysis
First of all the reliability of the scales was evaluated using Cronbachs alpha, which
measures the consistency with which respondents answer questions within a scale.
Regarding the answers to the research questions, some descriptive statistics were used.
Finally, regression analysis and one-way ANOVA were run to test the research
hypotheses.
Reliability of the measures
It was mentioned above that the questionnaire adopted in this study consists of 43
questions distributed over six aspects of risk. Reliability of the measures was assessed
with the use of Cronbachs alpha. Cronbachs alpha allows us to measure the reliability
of different variables. It consists of estimates of how much variation in scores of
different variables is attributable to chance or random errors (Selltiz et al., 1976). As a
general rule, a coefcient greater than or equal to 0.7 is considered acceptable and a
good indication of construct reliability (Nunnally, 1978). The overall Cronbachs alpha
for the six aspects is (0.70). Cronbachs alpha for the individual aspects
understanding risk and risk management (URM); risk identication (RI); risk
assessment and analysis (RAA); risk monitoring (RM); risk management practices
(RMP); and credit risk analysis (CRA) is (0.679), (0.689), (0.620), (0.652), (0.677) and
(0.638) respectively. These results show that all of these aspects are reliable (see
Table I).
Risk management aspects Cronbachs alpha
Understanding risk and risk management (URM) 0.679
Risk identication (RI) 0.689
Risk assessment and analysis (RAA) 0.620
Risk monitoring (RM) 0.652
Risk management practices (RM) 0.677
Credit risk analysis (CRA) 0.638
Table I.
The six risk management
aspects and their internal
consistency
Banks risk
management
399
Results of testing the research questions and hypotheses
In order to answer our research questions, the following risk management aspects will
be dealt with.
Understanding risk and risk management
Table II shows that the mean of responses on the eight questions about understanding
risk and risk management is 5.820. The respondents answers on these eight questions
indicate that the UAE banks staff understand risk and risk management, giving a
positive answer on the rst question. The table also indicates the relative importance of
the eight questions. However, there is not a big difference between the highest and lowest
means of the eight questions, as Table II shows. The highest mean (6.00) was that of
question four, in which respondents viewed managing risk as important to the
performance and success of their bank; the lowest mean (5.490) was of item eight,
concerning the banks objective regarding the application of advanced risk management
techniques, which is a feature of good risk management. It is obvious that the UAE
banks staff have a good understanding of risk and risk management, which might give
an indication about the ability of these banks to manage risks efciently in the future.
Risk identication
Risk identication is very important step in risk management. The study questionnaire
includes ve questions about risk identication. Table III provides the means and
standard deviations. It can be seen from the average of the means (5.062) that the UAE
banks have clearly identied the potential risks relating to their declared aims and
objectives. The answers on the ve questions about risk identication represent a
positive answer on question two of our research questions, which is also consistent
with the above-mentioned conclusion regarding risk understanding, as the more the
staff understand the risk, the more easily they can identify it. The table also indicates
Questions
Frequency of
5, 6 and 7 % Means SD
1 There is a common understanding of risk
management across the bank 146 91 5.7 0.758
2 Responsibility for risk management is clearly set out
and understood throughout the bank 141 89.8 5.8 0.841
3 Accountability for risk management is clearly set out
and understood throughout the bank 142 90.45 5.87 0.827
4 Managing risk is important to the performance and
success of the bank 145 7.7 6.00 1.085
5 It is crucial to apply the most sophisticated
techniques in risk management 137 12.7 5.834 1.309
6 Your banks objective is to expand the applications
of advanced risk management techniques 145 7.7 5.987 0.884
7 It is important for your bank to emphasize on the
continuous review and evaluation of the techniques
used in risk management. 147 6.4 5.885 1.091
8 Applications of risk management techniques reduce
costs or expected losses 133 84.7 5.490 1.089
Average 5.820
Table II.
Respondents answers on
questions about
understanding risk and
risk management
JRF
8,4
400
the relative importance of each question. The rst question obtained the highest mean
value of 5.636, indicating that the UAE banks carry out a comprehensive and
systematic identication of their risks, followed by question ve with a mean value of
5.554, which indicates that the UAE banks have developed and applied procedures for
the systematic identication of investment opportunities. However, the UAE banks
appear to face difculties in prioritizing their main risks, given the mean value of the
respondents answers on question two shown in Table III. In other words, the UAE
banks need to know how to prioritize their main risks efciently.
Regarding risk identication methods, the questionnaire includes a closed-ended
question about risk identication methods based on an ordinal scale, as previously
described. Table IV shows the respondents answers on this question. It can be seen
that the most important four methods, chosen by more than 90 per cent of the
respondents, are:
(1) inspection by the bank risk manager;
(2) audits or physical inspection;
(3) nancial statement analysis; and
(4) risk survey.
No. Questions
Frequency of
5, 6 and 7 % Means SD
1 The bank carries out a comprehensive and
systematic identication of its risks relating to each
of its declared aims and objectives 137 87.3 5.636 1.050
2 The bank nds it difcult to prioritize its main risks 62 60.6 4.172 1.637
3 Changes in risk are recognized and identied with
the banks roles and responsibilities 131 883.6 5.356 1.043
4 The bank is aware of the strengths and weaknesses
of the risk management systems of other banks 81 51.8 4.592 1.621
5 This bank has developed and applied procedures for
the systematic identication of investment
opportunities 138 87.9 5.554 0.901
Average 5.062
Table III.
Respondents answers on
questions about risk
identication
No. Risk identication method Frequency %
1 Inspection by the bank risk manager 155 98.7
2 Audits or physical inspection 149 94.9
3 Financial statement analysis 145 92.4
4 Risk survey 143 91.1
5 Process analysis 132 84.1
6 SWOT (strengths, weaknesses, opportunities, threats) analysis 130 82.8
7 Inspection by outside expert 130 82.8
8 Benchmarking 125 79.6
9 Scenario analysis 120 76.4
10 Internal communication, such as internal conversation with employees 110 70.1
11 Others 39 24.8
Table IV.
Frequency distribution of
risk identication
methods
Banks risk
management
401
These results are consistent with those reached by Al-Tamimi (2002). It can be
concluded that the UAE commercial banks are highly aware of risk identication.
Risk assessment and analysis
The questionnaire includes seven questions about risk assessment analysis. The results
of the responses are shown in Table V. The mean of the sample responses on the seven
questions is 5.650, which indicates that the UAE commercial banks are efciently
assessing and analyzing risk, representing a positive answer on the third of our research
questions. It can also be seen from the table that there is not a big difference between the
means of the seven questions, which means that respondents viewed fairly equally the
questions of risk assessment, such as the analysis of the likelihood of risks, the use of
quantitative analysis methods, the use of qualitative analysis methods, the assessment of
the costs and benets of addressing risks, and the prioritizing of risks.
Risk monitoring
Risk monitoring can be used to make sure that risk management practices are in line
with desired practices. Proper risk monitoring also helps bank management to discover
mistakes early. The questionnaire includes six questions addressing risk monitoring.
Table VI summarizes the sample responses on these questions. The mean of the
samples responses on the six questions is 5.76, which indicates that the UAE
commercial banks have an efcient risk monitoring and controlling system. The
samples responses represent a positive answer on the fourth question of our research
questions. It can also be seen from the table that there is no big difference between the
means of the six questions, which indicates that respondents viewed the questions of
risk monitoring fairly equally.
Risk management practices
Risk management practices might be considered the most important aspect of risk
management. Even if the bank staff understand risk and risk management, risks are
No. Question
Frequency of
5, 6 and 7 % Means SD
1 This bank assesses the likelihood of occurring. risks 136 86.7 5.407 0869
2 This banks risks are assessed by using quantitative
analysis methods 135 86 5.535 1.047
3 This banks risks are assessed by using qualitative
analysis methods(e.g. high, moderate, low) 130 82.9 5.535 1.047
4 Your bank analyses and evaluates opportunities it has
to achieve objectives 146 93 5.885 0.0986
5 Your banks response to analyzed risks includes an
assessment of the costs and benets of addressing risks 140 89.2 5.866 0.934
6 Your banks response to analyzed risks includes
prioritizing of risks and selecting those that need active
management 136 86.6 5.770 0.992
7 Your banks response to analyzed risks includes
prioritizing risk treatments where there are resource
constraints on risk treatment implementation 133 84.8 5.554 0.989
Average 5.650
Table V.
Respondents answers on
questions about risk
assessment and analysis
JRF
8,4
402
clearly identied and the bank adopts sophisticated methods in risk assessment and
risk analysis, it still may not be the case that there are efcient risk management
practices. The questionnaire includes ten questions about risk management practices;
Table VII shows the samples responses to be 5.50, which indicates that the UAE banks
are efcient in risk management practices. The highest value, 6.070, occurs on question
six, which states Efcient risk management practices is one of the banks objectives.
This also conrms that UAE banks intend to have efcient risk management. The
lowest value, 4.917, occurs on question eight, which states The application of Basel
Capital Accord by your bank would improve the efciency of risk management. The
answers give an indication that the UAE banks staff should be more aware of the
application of the Basel Capital Accord, which was introduced to improve the
efciency of banks risk management. This recommendation is mainly based on the
authors interviews with a large number of the respondents, as well as on the
respondents answers on question eight.
The questionnaire includes a general question about risk management practices:
Overall, I consider the level of risk management practices of this bank to be excellent.
The mean of the samples responses is 5.484, which supports the above arguments
regarding efcient risk management practices.
Credit risk analysis
It was mentioned earlier that credit risk of commercial banks represents the most
important type of risk, and also that the UAE banks did encounter a credit risk and
that they might face this type of risk again in the future. As credit risk is the most
important type of risk, the questionnaire includes seven questions about credit risk.
Table VIII provides information about the samples responses on these questions. The
mean of the responses on the seven questions is 5.443, which provides evidence about
the efciency of credit risk management in the UAE commercial banks. The most
important answers were on questions about undertaking credit worthiness analysis;
undertaking a specic analysis including the clients character, capacity, collateral
capital and conditions; and requiring sufcient collateral.
No. Question
Frequency of
5, 6 and 7 % Means SD
1 Monitoring the effectiveness of risk management is
an integral part of routine management reporting 136 86.6 5.719 1.018
2 The level of control by the bank is appropriate for the
risks that it faces 141 89.8 5.783 1.178
3 Reporting and communication processes within your
bank support the effective management of risk 141 89.8 5.802 1.070
4 The banks response to risk includes an evaluation of
the effectiveness of the existing controls and risk
management responses 141 89.8 5.789 0.0967
5 The banks response to risk includes action plans for
implementing decisions about identied risks 142 91.1 5.751 1.016
6 The banks response to risk includes an assessment
of the costs and benets of addressing risks 149 94.9 5.758 0.0983
Average 5.76
Table VI.
Respondents answers on
questions about risk
monitoring and analysis
Banks risk
management
403
No. Question
Frequency of
5, 6 and 7 % Means SD
1 This bank undertakes a credit worthiness analysis
before granting loans 144 91.8 5.719 1.164
2 Before granting loans your bank undertake a specic
analysis including the clients characters, capacity,
collateral capital and conditions 133 84.8 5.649 1.085
3 This banks borrowers are classied according to a
risk factor (risk rating) 130 83.4 5.592 1.037
4 It is essential to require sufcient collateral from the
small borrowers 127 80.9 5.343 1.259
5 This banks policy requires collateral for all granting
loans 114 57 5.133 1.177
6 It is preferable to require collateral against some
loans and not all of them 117 74.5 4.955 1.528
7 The level of credit granted to defaulted clients must
be reduced 17 89 5.713 1.121
Average 5.443
Table VIII.
Respondents answers on
questions addressing
credit risk analysis
Questions
Frequency of
5, 6 and 7 % Means SD
1 The banks executive management regularly reviews
the organizations performance in managing its
business risks 144 92.7 5.751 1.089
2 Your bank has highly effective continuous
review/feedback on risk management strategies and
performance 144 92.7 5.732 1.015
3 The banks risk management procedures and
processes are documented and provide guidance to
staff about managing risks 138 87.9 5.662 1.028
4 Your banks policy encourages training programs in
the area of risk management 143 91.1 5.815 0.0853
5 This bank emphasizes the recruitment of highly
qualied people in risk management 140 89.2 5.828 1.261
6 Efcient risk management is one of the banks
objectives 146 93 6.070 0.0994
7 It is too dangerous to concentrate banks funds in one
specic sector of the economy 89 56.8 4.961 1.628
8 The application of Basel capital Accord by your
bank would improve the efciency of risk
management 86 54.8 4.917 1.423
9 Banks capital is adequate if the ratio of capital to
total risk-weighted assets is equal to 8 percent 85 54.2 4.866 1.276
10 Overall, I consider the level of risk management
practices of this bank to be excellent 134 85.4 5.484 1.065
Average 5.508
Table VII.
Respondents answers on
questions addressing risk
management practices
JRF
8,4
404
Types of risk
The UAE commercial banks face different types of risk, and the respondents were asked
to indicate and rank in order of importance the risks facing their bank. Their answers are
summarized in Table IX. It was found that UAE commercial banks face all different
types of risk, but to varying degrees, the most important three types being foreign
exchange risk, followed by credit risk, then operating risk. The authors believes the
answers on this question were actually not accurate for two reasons: rst, as far as the
researches know the UAE banks do face foreign exchange risk but it is not the most
important risk; and secondly, the respondents ranked liquidity risk fourth, which is
questionable because the evidence indicates that the UAE banks do not suffer from
liquidity problems. For example, the liquidity ratio (i.e. total loans and advances divided
by total deposits) was 76 per cent in 2004, which can be interpreted to mean that the UAE
had sufcient liquidity (Emirates Banks Association, 2004), the other interpretation
being that the UAE banks did not utilize the available resources properly.
In order to test H1, the following regression model is used:
RMP f URM; RI; RAA; RM; CRA
where:
RMP risk management practices;
URM understanding risk and risk management;
RI risk identication;
RAA risk assessment and analysis;
RM risk monitoring; and
CRA credit risk analysis.
Using more than one variable to examine the contribution of independent variables to
the regression model may suggest a multicollinearity problem among these variables.
A multicollinearity test was carried out to assess the degree of correlation among
variables. Pearsons correlation was used to analyze correlations among the
independent variables, namely understanding risk and risk management (URM),
No. Type of risk Frequency %
1 Foreign exchange risk 152 96.82
2 Credit risk 151 96.18
3 Operating risk 147 93.63
4 Liquidity risk 141 89.81
5 Legal risk 141 89.81
6 Solvency risk 133 84.71
7 Interest rate risk 128 81.53
8 Counterparty risk 126 80.25
9 Price risk 120 76.43
10 Reputation risk 118 75.16
11 Strategic risk 117 74.52
12 Others 78 49.68
Table IX.
Types of risks facing the
UAE commercial banks
Banks risk
management
405
risk identication (RI), risk assessment and analysis (RAA), risk monitoring (RM), and
credit risk analysis (CRA). Table X reveals the correlation coefcients between the
independent variables. The rule of thumb test suggested by Anderson et al. (1990)
states that any correlation coefcient exceeding (0.7) indicates a potential problem. An
examination of the results of correlations presented in Table X suggests that there is no
problem of multicollinearity among the independent variables.
Table XI shows the regression results. It can be seen from the results provided in
Table XI that the R square is 0.235. This indicates that the ve independent variables
explain 23.5 per cent of the variations in risk management practices. The estimated
coefcients of two independent variables were, as expected, positive and statistically
signicant at the 1 per cent level in the case of RI and at the 5 per cent level in the case
of RAA. The estimated coefcients of URM, RM and CRA had insignicant positive
impact on risk management practices.
Hypothesis 1. It can be concluded that H1 of our research hypotheses is conrmed.
Furthermore, the results indicate that risk identication (RI) and risk assessment an
analysis (RAA) were the most important variables or the most inuencing variables in
risk management practices, which means that the UAE banks need to give more
attention to risk identication and risk assessment and analysis.
In order to test the remaining ve hypotheses, a one-way ANOVA was run. The
results of the analysis will be ordered according to the research hypotheses. Table XII
shows the results of ANOVA analysis.
Hypothesis 2. There is a difference between the UAE national and foreign banks in
the understanding of risk and risk management. Table XII reveals that there is a
signicant difference between the UAE national and foreign banks in the
understanding of risk and risk management, which means H2 is conrmed. These
results were expected because it is assumed that staff of foreign banks have better
URM RI RAA RM CRA
URM 1.000
RI 0.066 1.000
RAA 0.389
* *
0.273
* *
1.000
RM 0.437
* *
0.210
* *
0.461
* *
1.000
CRA 158
*
0.286
* *
0.269
* *
0.238
* *
1.000
Notes:
* *
Correlation is signicant at the 0.01 level (2-tailed);
*
correlation is signicant at the 0.05
level (2-tailed)
Table X.
The correlation
coefcients between
independent variables
R R
2
Adjusted R
2
F
0.484 0.235 0.209 9.255
Beta t Sig.
(Constant) 3.605 0.000
URM 0.271 0.160 0.873
RI 0.200 3.545 0.001
RAA 0.092 2.350 0.020
RM 0.120 1.071 0.286
CRA 0.013 1.570 0.119
Table XI.
OLS regression results
JRF
8,4
406
understanding of risk and risk management, compared with staff of the national
banks. Based on this nding, the UAE national banks should give more attention to
training and professional development of their staff in the area of risk management.
Hypothesis 3. There is a difference between the UAE national and foreign banks in the
practice of risk identication. Table XII shows that there is no signicant difference
between the UAE national and foreign banks in the practices of risk identication, which
means H3 is not conrmed. These results are consistent with those provided in Table III,
where it was concluded that all the UAE banks (i.e. both national and foreign banks) are
clearly identifying the potential risks relating to each of their declared aims and objectives.
Hypothesis 4. There is a difference between the UAE national and foreign banks in the
practice of risk assessment and analysis. Table XII indicates that there is a signicant
difference between the UAE national and foreign banks in the practice of risk assessment
and analysis, which means that H4 is conrmed. These results were expected because it
is assumed that foreign banks have more qualied staff, and accordingly they will be
more efcient in risk assessment and analysis. Thus, the UAE national banks need to
review the methods and techniques they use in risk assessment and analysis.
Hypothesis 5. There is a difference between the UAE national and foreign banks in
risk monitoring and controlling. Table XII reveals that there is a signicant difference
between the UAE national and foreign banks in risk monitoring and controlling, which
means that H5 is conrmed. These results also were expected, and are consistent with
the above-mentioned results, which indicate that there is a difference between the UAE
national and foreign banks in understanding risk and risk management and the
practices of risk assessment and analysis. The difference between UAE national and
foreign banks in risk monitoring and controlling is highly related to these two aspects.
In this regard, the UAE national banks need to review the methods and techniques
they use in monitoring and controlling risk activities.
Hypothesis 6. There is a difference between the UAE national and foreign banks in
the context of risk management practices. Table XII shows that there is no signicant
Source SS DF MS F-value Sig.
URM Between groups 0.970 1 0.970 3.970 0.048
Within groups 37.870 155 0.244
Total 38.840 156
RI Between groups 0.686 1 0.686 1.276 0.260
Within groups 83.382 155 0.538
Total 84.068 156
RAA Between groups 2.960 1 2.960 7.762 0.006
Within groups 59.097 155 0.381
Total 62.057 156
RM Between groups 2.498 1 2.498 5.690 0.018
Within groups 68.044 155 0.439
Total 70.542 156
RMP Between groups 0.485 1 0.485 1.523 0.219
Within groups 49.402 155 0.319
Total 49.888 156
CRA Between groups 0.208 1 0.208 0.739 0.391
Within groups 43.571 155 0.281
Total 43.779 156
Table XII.
The results of analysis of
variance for national
banks and foreign banks
Banks risk
management
407
difference between the UAE national and foreign banks in the context of risk
management practices, which means that H6 is not conrmed. These results were not
expected and are not consistent with those mentioned above. The results might be
attributed to the fact that both the national and foreign banks are working in the same
environment and subject to same rules and instructions.
Hypothesis 7. There is a difference between the UAE national and foreign banks in
the practice of credit risk analysis. Table XII reveals that there is no signicant
difference between the UAE national and foreign banks in the practice of credit risk
analysis, which means that H7 is not conrmed. These results were not expected and are
not consistent with the above-mentioned results, for the same reason mentioned above.
Conclusions
The main results of this study are:
(1) The three most important types of risks facing the UAE commercial banks are
foreign exchange risk, followed by credit risk, then operating risk.
(2) The four most important methods of risk identication, chosen by more than 90
per cent of the respondents, are: inspection by the bank risk manager, audits or
physical inspection, nancial statement analysis and risk survey.
(3) The UAE banks are somewhat efcient in assessing and analyzing risks, risk
management practices, risk monitoring and, and risk identication.
The results also provide some amount of evidence about efcient credit risk
management in the UAE commercial banks. Finally, the results indicate that there is a
signicant difference between the UAE national and foreign banks in the practice of
risk assessment and analysis and in risk monitoring and controlling, whereas there is
no signicant difference between the UAE national and foreign banks in the aspects of
risk identication, risk management practices and the practice of credit risk analysis.
Areas for further research
The authors tried in this study to cover most of the aspects of risk management.
However, this paper did not address in detail credit risk management. This type of risk
can be addressed in future studies as credit risk represents the most challenging type of
risk. Further research may also consider analyzing liquidity risk management as
liquidity position affects the continuity of commercial banks and a weak liquidity
position might lead to the liquidation of commercial banks. Further research may also
focus on Basel II and risk management, one of the hottest topics in the banking industry.
Finally, the study could usefully be conducted in another country, using the same
methodology. Different and interesting results may be expected, because risk
management practices are mainly affected by specic factors such as economic
conditions, competition and regulations.
References
Al-Tamimi, H. (2002), Risk management practices: an empirical analysis of the UAE commercial
banks, Finance India, Vol. 16 No. 3, pp. 1045-57.
Anderson, R.A., Sweeney, D.J. and Williams, T.A. (1990), Statistics for Business and Economics,
West Publishing Company, St. Paul, MN.
JRF
8,4
408
Basel Committee on Banking Supervision (1999), Principles for the management of credit risk,
consultative paper issued by the Basel Committee on Banking Supervision, issued for
comment by 30 November, p. 3.
Carey, A. (2001), Effective risk management in nancial institutions: the Turnbull approach,
Balance Sheet, Vol. 9 No. 3, pp. 24-7.
Devellis, R.F.. (1991), Scale development: theory and application, Applied Research Methods
Service, Vol. 26, Sage Publications, Newbury Park, CA.
Emirates Banks Associations (2004), Financial position of commercial banks in the UAE,
Emirates Banks Associations, Abu Dhabi.
Hahm, J.H. (2004), Interest rate and exchange rate exposures of banking institutions in pre-crisis
Korea, Applied Economics, Vol. 36 No. 13, pp. 1409-19.
Khambata, D. and Bagdi, R.R. (2003), Off-balance-sheet credit risk of the top 20 Japanese banks,
Journal of International Banking Regulation, Vol. 5 No. 1, pp. 57-71.
Linbo Fan, L. (2004), Efciency versus risk in large domestic US, Managerial Finance, Vol. 30
No. 9, pp. 1-19.
Niinimaki, J.-P. (2004), The effects of competition on banks risk taking, Journal of Economics,
Vol. 81 No. 3, pp. 199-222.
Nunnally, C.J. (1978), Psychometric Theory, McGraw-Hill, New York, NY.
Oldeld, G.S. and Santomero, A.M. (1997), Risk management in nancial institutions, Sloan
Management Review, Vol. 39 No. 1, pp. 33-46.
Salas, V. and Saurina, J. (2002), Credit risk in two institutional regimes: Spanish commercial and
savings banks, The Journal of Financial Services Research, Vol. 22 No. 3, pp. 203-16.
Selltiz, C., Wrightsman, L.S. and Cook, W. (1976), Research Methods in Social Relations, Holt,
Rinehart and Winston, New York, NY.
Wang, A.T. and Sheng-Yung, Y. (2004), Foreign exchange risk, world diversication and
Taiwanese ADRs, Applied Economics Letters, Vol. 11 No. 12, pp. 755-8.
Wetmore, J.L. (2004), Panel data, liquidity risk, and increasing loans-to-cor deposits ratio of large
commercial bank holding companies, American Business Review, Vol. 22 No. 2, pp. 99-107.
Further reading
Bank for International Settlements (1999), Consultative paper on a new capital adequacy
framework, press releases.
Corresponding author
Hussein A. Hassan Al-Tamimi can be contacted at: husseintam@hotmail.com
Banks risk
management
409
To purchase reprints of this article please e-mail: reprints@emeraldinsight.com
Or visit our web site for further details: www.emeraldinsight.com/reprints
The North Cyprus banking sector:
the effect of a speculative attack
on the Turkish Lira
Nil Gunsel
Near East University, Nicosia, North Cyprus, Cyprus
Abstract
Purpose The purpose of this research is to investigate the effect of a speculative attack on the
Turkish Lira in the North Cyprus banking sector during the period 1984-2002.
Design/methodology/approach A mutivariate logit model is the empirical methodology
employed in this analysis that allows us to identify the determinants of the probability of bank
failure. In the model, the existence of contagious currency crises is constructed as an index of exchange
market pressure, which is a weighted average of changes in interest rates, international reserves and
the nominal exchange rate.
Findings The empirical result reveals that the a speculative attack on the Turkish Lira in 1994 and
2001 put stress on banks operating in North Cyprus and led to banking sector distress. The ndings
also suggest that bank-specic weaknesses, high interest rates, high credit, low trade and the xed
exchange rate policy signicantly increased the bank fragility.
Research implications/limitations For further research this paper may better distinguish
contagion if it uses economic and nancial ties from Turkey that are practically susceptible to bank
failure in North Cyprus.
Practical implications This paper presents a practical application of a currency crisis model in
the North Cyprus banking sector. In addition to the risk of currency crises, risk under xed rate
regimes, interest rate risk, trade risks and credit risk are also used to encourage correct risk
management behaviour in the North Cyprus banking sector.
Originality/value This analysis would appear to be the rst systematic evidence that investigates
the effect of a speculative pressure on Turkish Lira in the North Cyprus banking sector.
Keywords Cyprus, Turkey, Banking, Foreign exchange, Market forces, Financial risk
Paper type Research paper
Introduction
After the division of Cyprus in 1974, the North Cyprus Government started to be
persistently, economically and nancially dependent on Turkey. As the ofcial
currency in North Cyprus is Turkish Lira, the monetary policy of the Turkish Cypriots
tends to be geared to the economy of Turkey, with little reference to the situation in
North Cyprus. In this sense, due to the unstable economic situation in Turkey, the
North Cyprus economy has also suffered from the continuous deterioration of
macroeconomic fundamentals (such as destabilizing, rapid devaluation and high
ination) for many years.
The North Cyprus economy has experienced two banking sector distress periods.
The rst took place in 1994 and the second took place between the years 2000 and
2002. In early 1994, the economic fundamentals in Turkey were deteriorating.
Particularly, there was a continuing devaluation of the TL, which resulted in a
serious currency crisis. As there is a close monetary and economic link between
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1526-5943.htm
JRF
8,4
410
The Journal of Risk Finance
Vol. 8 No. 4, 2007
pp. 410-421
qEmerald Group Publishing Limited
1526-5943
DOI 10.1108/15265940710777342
Turkey and North Cyprus, as a consequence of the nancial distress experienced in
Turkey in 1994, banks in North Cyprus were also affected. In 1994 two banks,
namely Everest Bank Ltd and Mediterranean Guarantee Bank Ltd, were placed
under the control of the TRNC Ministry of Finance. Later, these banks had to be
bailed out by the Government.
The North Cyprus economy also experienced severe economic and nancial
problems between the years 2000-2002. In particular, the onset of the Turkish
currency crises during 2000/2001 appears to have had a negative impact on the NC
banking sector and led to contraction of the economy. In December 1999, the
International Monetary Fund (IMF) supported the pegged exchange rate base
anti-ination programme implemented in Turkey. However, after fourteen months
from the start of the programme, i.e. in February 2001, the programme had to be
abandoned with the collapse of the TL. During the period of 2000-2002 ten nancial
banks were forced by the Government of North Cyprus to suspend their operation.
In 2000, ve banks, namely the Cyprus Credit Bank Ltd, Cyprus Liberal Bank Ltd,
Everest Bank Ltd, Kibris Yurtbank Ltd and Cyprus Finance Bank Ltd, were put
under the Saving Deposit Insurance Fund (SDIF), and then these banks were closed
in the year 2001. The bankruptcy of these ve banks started a serious banking
crisis in North Cyprus. Another four banks, namely Cyprus Commercial Bank Ltd,
Yasa Bank Ltd, Tilmo Bank Ltd and Asia Bank Ltd, were put under the SDIF in
2001, and Cyprus Industrial Bank Ltd was put under the SDIF in 2002.
Furthermore, Finba Ltd was taken over by Artam Bank Ltd in 2000 and Med Bank
Ltd and Hamza Bank Ltd were taken over by Seker Bank Ltd in the years 2001 and
2002 respectively. During 1999 there were 37 surviving banks in North Cyprus.
However, towards the end of 2002 ten of these banks suspended operations, two
banks were taken over by another bank, and only 25 banks remained. The increase
in the failure of commercial banks in North Cyprus increased attention on efforts to
investigate the determinants of bank failure.
Review of literature
The last two decades have witnessed an unprecedented increase in the number of
nancial distress episodes, both in developed and developing countries. Therefore, the
issue of analyzing the determinants of the nancial crises have become increasingly
important for economies. In general, the nancial problems that have received the most
attention in the literature are currency and banking crises. For instance, in the early
1980 s, several Latin American countries including Mexico, Brazil and Chile
experienced a banking crisis. In particular, the 1990 s were marked by an unusual
number of nancial and economic crises such as the attack of the European Exchange
Rate Mechanism in 1992-1993, the Mexican tequila crisis in 1994-1995, the Turkey
currency and banking crisis in 1994, the Asian nancial crisis affecting Korea,
Thailand, Malaysia, and Indonesia in 1997-1998, the Russian default in 1998 and its
spill-over into Latin America (the Brazilian crises in 1998-1999), the Turkish currency
and banking crisis in 2000-2001 and the difculties concerning the Argentinean
economy in 2001-2002. The attack on the Brazilian real in 2002 reminded the world
how rapidly and unexpectedly a nancial crisis can erupt, causing disruption in credit
channels, economic contraction and spill-over to other countries and how difcult it can
be to provide an effective policy response. (see for instance, Lindgren et al. (1996);
The North
Cyprus banking
sector
411
Hutchison and McDill (1999); Kaminsky and Reinhart (1998); Saxena (2004)). These
events increased concerns about banking and currency crises and led economists to
focus on the causes of these crises.
Theoretical models of crises can be examined under three categories:
rst-generation models, second-generation models and third-generation models. The
rst-generation models focus on the role of weak fundamentals in the policy of
government as a triggering factor of currency crises. The scal decit, growth of
money supply, current account balance and the level of foreign exchange reserves are
the economic indicators that are derived from this framework. This model assumes
that the government budget decit is the root of speculative attacks on pegged
exchange rates. In general, the main features of the crisis in Latin America in the early
1980 s can be dened by rst-generation models. (see Krugman (1979); Saxena (2004)).
However, the rst-generation models failed to explain the crises in Europe
(1992-1993) and in Mexico (1994-1995). This failure led to the development of
second-generation models. The second-generation currency crises models offered no
critical developments to fundamental macroeconomic variables; instead it provided
model of self-fullled speculative attacks in foreign exchange markets. In this model
market expectations directly inuence macroeconomic policy decisions in a rational
direction (see Obstfeld (1986)).
When nancial crises erupted in South East Asia in 1997, neither rst nor
second-generation models explained the reasons for failure. During this crisis period in
Asian countries ination, government decit and unemployment were low and credit
was high. Hence, this crisis in Asia has revealed the need for a new framework that
integrates weaknesses in the banking sector into the early generation models. This new
framework, which takes into account the weaknesses of the nancial sector, is called
the third- generation models or the term twin crises, which emphasize the occurrence of
both banking and currency crises. (See Krugman (1999); Radelet and Sachs (1998);
Kaminsky and Reinhart (1998).
The theoretical framework of third-generation models is the analysis of nancial
crises that combine two important principles (weak macroeconomic fundamental and
self-fullled attack) of rst and second-generation models with the banking sector. In
addition, this new framework considers new issues such as moral hazards, the herding
behaviour of bankers and portfolio managers and international contagion effects
appearing as a result of trade or nancial linkage between countries. In this context,
the nancial fragility-based explanation relies on the third-generation models.
Explanatory variables
Gonzalez-Hermosillo (1996) dened bank fragility as a function of bank-specic factors
that are common to individual banks and macro factors that are common to all banking
sectors. From this perspective, in this paper it is argued that the macro environment
and potential contagion currency crises from Turkey may exacerbate the internal
troubles of many of the nancial institutions in North Cyprus. Therefore, the model
showing the fragility of bank i may be expressed as a function of:
F
i
f
i
Micro; Macro; Contagion 1
where:
JRF
8,4
412
Micro bank-specic risk in the context of CAMELS criteria;
Macro the risk of a macro-environment, which represents weaknesses in the
economy and nancial system; and
Contagion the contagion effect from Turkey.
In this framework, it is assumed that basically the bank fragility in North Cyprus
depends on a micro approach (bank-specic risks), macro-environment and potential
contagion effect from Turkey. Descriptions and expected signs of macro variables are
illustrated in Table I.
A number of microeconomic (bank-specic) factors and macro factors are
considered by theory as good indicators of banking fragility. Bank-specic variables
are the key nancial ratios that are designed to reect CAMELS components. The
CAMELS approach appears to be appropriate for identifying weaknesses specic to
individual banks. These are ratios that reect capital adequacy (total capital as a
percentage of total assets), asset quality (total loans as a percentage of total assets),
earning (net income as a percentage of total assets), liquidity (total deposits as a
percentage of total loans) and asset size (total sectoral asset as a percentage of total
assets). Furthermore, four sets of macro variables, namely macroeconomic variables,
nancial variables, external conditions, as well as potential contagion effect from
Turkey are also included in the model.
Methodology
Analyzing the contagion effect from Turkey
The contagion effect that is the concern of this paper is an increase in the
probability of speculative attacks on the Turkish Lira. Currency crises are
Variable name Denition Expected sign failure
Bank specic variables
Capital/asset Total capital as a percentage of total assets
Loan/asset Total loans as a percentage of total assets
Net income/asset Net income as a percentage of total assets
Deposit/loan Total deposit as a percentage of total loans 2 /
Asset size (1) Total sectoral asset as a percentage of total assets
Macroeconomic variables
Real interest rate (RIR) The real interest rates
Financial variables
Private credit The ration of private domestic credit to GDP
Public credit The ratio of the public credit to GDP
External conditions
Exchange rate regime Dummy for pegged exchange regime 2 /
Terms of trade (TT) The terms of trade change
a

Contagion effect
Market pressure The market pressure index in Turkey
Note:
a
Changes in the price of export over import
Table I.
Denition and expected
signs of macro variables
The North
Cyprus banking
sector
413
identied as extreme values of the speculative pressure index. Following
Eichengreen et al. (1996), a measure of speculative pressure on currency crises
(exchange rate pressure index) is constructed as a weighted average of changes in
the exchange rate, changes in the international reserve and changes in the interest
rate.
To examine currency crises in Turkey, an index of the weighted average of changes
in the exchange rates, foreign exchange reserves and interest rates are calculated.
Then, the Market Pressure Index (index of exchange rate pressure on Turkish Lira) is
calculated as follows (see Eichengreen et al., 1996):
MPI
i;t
a%De
i;t
bDi
i;t
2g%Dr
i;t
2
where:
e the nominal exchange rate vis-a`-vis the USA;
i
i,t
short-term interest rates;
r foreign exchange reserves; and
a, b, g weights.
A higher index is reected in higher values of these three variables; therefore, this
indicates greater pressure on the exchange market depending on the nature of the
intervention of the respective Central Bank. That is, speculative pressures are either
accommodated by a loss of reserves or can be prevented by the monetary authorities
through an increase in interest rates (see for instance, Eichengreen et al. (1996); Frankel
and Rose (1996); Cramazza et al. (2004)).
The logit model
This paper uses a logit model in a panel data framework. The one-step ahead
probability of failure is estimated as a function of a set of micro and macro
variables. From this perspective, before running the estimation techniques the
panel considers potentially combining time-series and cross section data. Using
pooled time series cross-section data (panel data) it is convenient to measure the
sensitivity of bank-specic data, macro data and contagion effects over time for
each bank.
In the context of the logit model, the binary dependent variable Y
it
takes the value of
1 if a bank fails (transferred to the SDIF, closed or taken over by another bank) during
the year, and 0 otherwise.
In practice, Y
it
*
is the latent variable, which is not observable by the researcher and
assumed to depend on k explanatory variables, ranging from 21 to 1. The latent
variable is linked to the observable Y
i
variable by a measurement equation.
The latent variable Y
it
*
is linked to the observable categorical variable as follows (see
Maddla (2001, p. 322):
Y
it

1
0
If individual banks fail If Y
*
it
. 0
otherwise If Y
*
it
# 0
2
4
3
JRF
8,4
414
The latent variable link to the explanatory variables as follows:
Y
it
b
0

X
k
j1
b
j
X
itj
u
it
4
where:
Y
it
*
latent variable, and its scale can not be determined;
u
it
composite error term;
b
j
coefcient of jth independent variable, and measures the effects on the odds
of failure of a unit change in the corresponding independent variables; and
X
itj
vector of k number of explanatory variables in period t for bank i.
The above equation implies that the larger values of Y
it
*
are observed as Y
it
1 (i.e.
failed banks), while those with smaller values of Y
it
*
are observed as Y
it
0 (i.e.
non-failed banks).
In the logit model, the log-odds ratio is a linear function of the explanatory variables
(see Maddla (2001)). The estimated multivariate logit model links the likelihood of
banking problems to a set of variables:
log
P
i
1 2P
i

b
o

X
k
j1
b
j
X
itj
5
where:
P
i
represents the probability that bank i will fail; and
1-P
i
represents the probability that bank i will not fail.
The empirical results
The univariate analysis
Including highly correlated macro variables in a model could result in signicant bias
to the level of the parameters. For this reason, in order to prevent the collinearity
problem, the models formed in this research do not include correlated variables in the
same model. In addition, in order to have a consistency between both micro and macro
data and to prevent the results from being spurious, we tested the stationarity property
by applying the Dickey-Fuller Unit Root Test. The results reveal that both micro and
macro data are stationary.
The multivariate analysis
The logit analysis. Table II presents the results of the multivariate logit model, which
estimates how a particular macro variable changes the probability of the occurrence of
the event when all other micro variables are constant. The logit regression results are
obtained with the use of STATA 8 Software. The number of observations range from
218 to 223, depending on data availability. The quality of model specication is
assessed based on the criteria of the Model Chi-square, Pseudo R
2
and AIC criteria. In
the model, the pseudo R
2
ranges from 56.20 to 63.17, implying that most of the
variation of the dependent variable is explained by the model explanatory variables. A
The North
Cyprus banking
sector
415
chi-square likelihood ratio test of the signicance of the overall model indicates that all
of the specication models are highly signicant (Prob . Chi-square: 0.000). The AIC
compares the model with different degrees of freedom and eliminates the model
specication when irrelevant explanatory variables are added into the regression. A
model with a lower value of the AIC is judged to be preferable.
With regard to bank-specic variables the rst ndings suggest that the coefcient
on the measure of capital adequacy is negative and statistically signicant, indicating
that the lower a banks shareholders equity to total assets ratio, the higher the
probability that the bank will fail in North Cyprus. The results suggest that the ability
of banks to absorb the cost of domestic and external shocks was seriously limited due
to being undercapitalized. At that time, regulations were allowing banks to work under
low capital requirements, which also contributed to the problem. This result is
consistent with Martin (1977), Avery and Hanweck (1984), Heffernan (1996), Langrin
(2001) and Yilmaz (2003).
Since loans are generally the most risky assets that banks hold, the ndings on
asset quality (ratio of loans to total assets) indicate that banks with high leverage are
more likely to fail. For most of the specication this result is statistically signicant at
1 per cent. After the bank failures it was noted that some of the failed banks in North
(1) (2) (3)
Variables N SD N SD N SD
Bank-specic variables
Capital/asset 20.054
*
0.028 20.062
*
0.034 20.064
* *
0.028
Loan/asset 0.017
* * *
0.006 0.017
* *
0.007 0.019
* *
0.006
Net income/asset 20.204
* * *
0.060 20.195
* * *
0.071 20.163
* *
0.072
Deposit/Loan 20.055
* *
0.023 20.057
* *
0.024 20.052
* * *
0.017
Asset size (1) 20.890
* * *
0.324 20.775
* *
0.311 20.894
* *
0.366
Macro variables
Real interest rates (Lag 1) 0.120
* * *
0.028 0.257
* * *
0.073 0.167
* *
0.075
Financial variables
Private credit (Lag 1) 2.009
* * *
0.649 3.449
* * *
1.084
Public credit (Lag 1) 4.912
*
2.555
Structural variables
Fixed exchange regime 3.392
* *
1.447
External conditions
Terms of trade 20.162
*
0.084
Contagion effect
Market pressure 0.848
* * *
0.382 0.808
*
0.374 0.642
* * *
0.413
Constant 21.825 1.329 0.147 2.919 25.703 2.050
Model statistics
Wald Chi2 95.24
* * *
74.74
* * *
65.74
* * *
Pseudo R
2
58.12 56.20 63.17
Log pseudo-lik 219.337 220.222 217.006
AIC 27.337 29.222 26.006
Notes:
*
,
* *
,
* * *
Indicates signicance at the 10, 5 and 1 percent level respectively; standard errors
are given in parentheses for the logit model; specication from 1-3 is the bank probability of
intervention model
Table II.
Logit analysis of
determinants of bank
intervention (contagion
effect from Turkey)
JRF
8,4
416
Cyprus were using their banks to channel funds into their own mostly unprotable
trading activities. Safakli (2003) stated that the amount of connected lending in the
failed banks ranged from 50 per cent to 90 per cent of total loans, which increased the
credit risk in those failed banks. The ndings imply that as banks were operating
under a poor credit and risk management, this may increase the possibility of poor
quality lending and lead to a bank failure. This result is consistent with the previous
ndings of Avery and Hanweck (1984), Thomson (1991) and Borovikova (2000).
The results on the protability measures (the ratio of net income to total assets) are
negative and for most specications it is statistically signicant at 1 per cent level,
which means an increase in the ratio of net income to total assets decreases the
probability of failure. These results agree with those obtained by Martin (1977), Avery
and Hanweck (1984), Thomson (1991), Heffernan (1996) and Persons (1999).
The ndings on the liquidity risk (the ratio of deposits to total loans) are negative
and for most specications they are statistically signicant at 1 per cent, suggesting
that a decrease in liquidity increases the probability of bank failure in North Cyprus. A
decrease in this ratio indicates that banks are illiquid to face unexpected deposit runs.
Safakli (2003) states that a banking crisis in North Cyprus was triggered by the failure
of Yurtbank A.S (located in Turkey) in December 1999. A major share of Kibris
Yurtbank Ltd in North Cyprus belonged to the owner of Yurtbank A.S. The failure of
Yurtbank A.S. in Turkey triggered a panic in North Cyprus, and depositors began to
withdraw funds from Yurtbank Ltd. The bank run quickly spilled over to other banks
in North Cyprus. By the end of January 2000, four illiquid banks that failed to face
unexpected deposit runs were put under the control of the Ministry of Finance.
Gonzalez-Hermosillo (1999) and Langrin (2001) also found that this ratio was
negatively related to the probability of failure, but their results were insignicant.
Finally, the measure of bank size (the ratio of total assets of a bank to total banking
sector assets) is negative and for most of the specication statistically signicant at 99
per cent condence level. This suggests that larger banks experienced a lower
probability of failure than smaller banks. The result suggests that large banks are
perceived to be too-big-to fail. Hence, the probability of failure is low.
Macroeconomic characteristics
In the relevant literature it is frequently argued that real interest rates have a lagged
effect. Following the studies of Gonzalez-Hermosillo et al. (1996) and Hardy Hardy and
Pazarbasoglu (1998) in this model the real interest rates are also considered to have a
lagged effect. The real interest rate is lagged by one-period, and as expected results
seem to be positive and statistically signicant at 1 per cent level. The one period
lagged suggests that the effect of real interest rates on bank balance sheet is not very
quick. As expected, the ndings support the view that an increase in the real interest
rate in the past increased the probability of bank failure in North Cyprus. These results
are in agreement with those obtained by Hardy and Pazarbasioglu (1998) and
Demirguc-Kunt and Detragiache (2000). This is probably due to the fact that an
unanticipated increase in interest rates makes it difcult for rms or households to
service their existing loans. As a result, this lead to an increase in banks
non-performing loans, which increases the probability of banking problems and
insolvency.
The North
Cyprus banking
sector
417
Financial variables
According to Sachs et al. (1996a) an increase in bank lending may not be a bad sign as
it shows that the economy is expanding. These authors argue that a growth of bank
credit to the private sector may be a cause of concern as this may lead to a decline in the
quality of credit. The ndings in the study reveal that one-year lagged of this variable
is positive and statistically signicant at 1 per cent level, which suggest that credit
extended to the private sector in the past increased the probability of bank failure in
North Cyprus. This implies that a lending boom can lead to a banking crisis, especially
if banks do not have good quality credit. This result is consistent with those obtained
by Demirguc-Kunt and Detragiache (1998a) in a sample of countries. Domac and
Martinez-Peria (2003), on the other hand argue that the ratio of the domestic credit to
the private sector to GDP can be used to capture the extent of nancial liberalization. In
this regard it can also be concluded that nancial liberalization increases the likelihood
of bank failure in North Cyprus.
The ndings suggest that one-year lagged of the credit that is extended to the public
sector has a positive sign and results seem to be statistically signicant at 5 per cent
level. The one period lagged implies that the effect of public credit on bank balance
sheets is not very quick. This suggests that an increase in the ratio of credit to public
sector to GDP in the past raised the likelihood of banking problems in North Cyprus.
The North Cyprus Government had limited reserves, and therefore, in order to pay the
salaries of the public workers, it was taking a large amount of domestic credit from the
Central Bank of North Cyprus. This suggests that the vulnerability of the banking
sector and the Governments inability to rescue nancial institutions may have
triggered the necessary conditions for a banking crisis in the year 2000-2001.
External conditions
The ndings illustrate that adopting a xed exchange rate regime is positive and
statistically signicant at 5 per cent level, suggesting that a xed exchange rate regime
increased the probability of failure in North Cyprus. This suggests that after the
implementation of a three year exchange rate based implementation programme in
December 1999 (pegged exchange rate was put into practice in Turkey and parallel to
this in North Cyprus), banks and rms in North Cyprus dominated their borrowing in
foreign currency without hedging foreign exchange rate risks. However, as a
consequence of the unexpected depreciation of the TL in February 2001 (14 months
from the start of the programme the currency peg had to be abandoned), this increased
the failure risk in banks and rms whose exchange rate was dominated by foreign
currency. The results of Domac and Martinez-Peria (2003), on the other hand, reveal
that adopting a xed exchange rate regime diminishes the probability of banking
crises. They found evidence that a xed exchange rate provides implicit guarantees for
investors that borrow in a foreign currency.
Furthermore, consistent with the results of Demirguc-Kunt and Detragiache (1998a)
and Hardy and Pazarbasioglu (1998) results in this study reveal that at 1 per cent and 5
per cent signicant level, adverse trade shocks increase the probability of bank
distress. As stated in the previous chapter, on the 5th of July 1994 the European Court
of Justice ruled that the European Union member states should not accept agricultural
product exported by the unrecognised area of the TRNC (Turkish Republic of Northern
Cyprus). This imposed an embargo on agricultural products, which was the major
JRF
8,4
418
export activity of the North Cyprus economy. The deterioration in the terms of trade
negatively affected the ability of borrowers (especially the ability of traders to pay their
debts) to repay loans, which resulted in the deterioration in bank balance sheets, thus
threatening the solvency of domestic banks and increased the probability of banking
sector problems in North Cyprus.
The contagion effect from Turkey
As the North Cyprus nancial market is highly integrated with Turkeys capital
market, it has been increasingly subject to the effects of changes in Turkey. The results
conrm that a speculative attack in Turkey is statistically signicant and increases the
vulnerability of the North Cyprus banking sector, even after controlling the economic
and nancial fundamentals in the country concerned. In other words, the exchange rate
pressure in Turkey in 1994 and 2001 put stress on banks that operate in North Cyprus
and led to banking sector distress. This would appear to be the rst systematic
evidence of the existence of contagious currency crises in North Cyprus.
Conclusions
The main inferences that can be drawn from the results of the empirical research
carried out in this paper can be summarized as follows:
According to the bank-specic determinants of bank failure capital inadequacy, low
asset quality, low protability, low liquidity and small size increased the risk of bank
failure. The ndings also suggest that high interest rates, high credit risk and low
trade signicantly increased the bank fragility in North Cyprus.
Further, the contagion effect from Turkey increased the vulnerability of the North
Cyprus nancial system. The ndings suggest that the exchange rate pressure in
Turkey (i.e. the speculative attack on the Turkish Lira) in 1994 and 2001 put stress on
banks operating in North Cyprus and led to banking sector distress. The results also
imply that the xed exchange rate policy signicantly increased the bank fragility in
North Cyprus. As noted earlier, in December 1999 Turkey had launched a three-year
IMF supported pegged exchange rate based stabilization program (xed exchange rate
regime) aimed to bring down ination. The new policy initially increased the incentive
of customers and banks to borrow in foreign currency. In other words, act as an
implicit guarantee against loss, which might be due to the unpredictability of the
exchange rate. However, 14 months after the start of the programme, in mid-February
2001, the currency peg had to be abandoned and replaced by a free-oating regime,
which led to an increase in the risk of bank failure in North Cyprus.
References
Avery, R.B. and Hanweck, G.A. (1984), A dynamic analysis of bank failures, Proceedings of the
20th Annual Conference on Bank Structure and Competition, Federal Reserve Bank of
Chicago, Chicago, IL, pp. 380-95.
Borovikova, V. (2000), The determinants of bank failures: the case of Belarus, MA thesis,
National University, Kiev, Mohyla Academy, pp. 1-22.
Cramazza, F., Ricci, L. and Salgado, R. (2004), International nancial contagion in currency
crises, Journal of International Money and Finance, Vol. 23, pp. 51-70.
Demirguc-Kunt, A. and Detragiache, E. (1998a), The determinants of banking crises in
developing and developed countries, IMF Staff Papers, Vol. 45, pp. 81-109.
The North
Cyprus banking
sector
419
Demirguc-Kunt, A. and Detragiache, E. (2000), Monitoring banking sector fragility: a
multivariate logit approach, The World Bank Economic Review, Vol. 14 No. 2, pp. 287-307.
Domac, I. and Martinez-Peria, M.S. (2003), Banking crisis and exchange rate regimes: is there a
link?, Journal of International Economics, Vol. 61, pp. 41-72.
Eichengreen, B., Rose, A. and Wyplosz, C. (1996), Contagious currency crises: rst tests,
Scandinanvian Journal of Economics, Vol. 98 No. 4, pp. 463-84.
Frankel, J.A. and Rose, A.K. (1996), Currency crashes in emerging markets: an empirical
treatment, Journal of International Economics, No. 41, pp. 351-66.
Gonzalez-Hermosillo, B. (1996), Banking sector fragility and systemic sources of fragility, IMF
Working Paper, No. 12, pp. 1-38.
Gonzalez-Hermosillo, B. (1999), Determinants of ex-ante banking system distress: a macro-micro
empirical exploration of some recent episodes, IMF Working Paper, No. 33, pp. 1-114.
Gonzalez-Hermosillo, B., Pazarbasioglu, C. and Billings, R. (1996), Banking system fragility:
likelihood versus timing of failure an application to the Mexican nancial crisis,
Working Paper No. 142, IMF.
Hardy, D.C. and Pazarbasioglu, C. (1998), Leading indicators of banking crises: was Asia
different?, Working Paper No. 91, IMF.
Heffernan, S. (1996), Modern Banking in Theory and Practice, John Wiley and Sons, Chichester,
pp. 268-308.
Hutchison, M. and McDill, K. (1999), Are all banking crises alike? The Japanese experience in
international comparison, Journal of Japanese and International Economics, No. 13,
pp. 155-80.
Kaminsky, G. and Reinhart, C. (1998), On crises, contagion, and confusion, Journal of
International Economics, No. 51, pp. 145-68.
Krugman, P. (1979), A model of balance-of-payments crisis, Journal of Money, Credit and
Banking, Vol. 11 No. 3, pp. 311-25.
Krugman, P. (1999), Currency Crisis, International Capital Flows, NBER and Chicago University
Press, Chicago, IL, pp. 421-40.
Langrin, R.B. (2001), The determinants of banking system distress: a microeconomic and
macroeconomic empirical examination of the recent Jamaican banking crisis, PhD thesis,
Pennsylvania State University, Harrisburg, PA, pp. 1-144.
Lindgren, C., Garcia, G. and Saal, M.I. (1996), Bank soundness and macroeconomic policy, IMF
Working Paper.
Maddla, G.S. (2001), Introduction to Economics, 3rd ed., John Wiley and Sons, Chichester.
Martin, D. (1977), Early warning of bank failure: a logit regression approach, Journal of
Banking and Finance, No. 1, pp. 249-76.
Obstfeld, M. (1986), Rational and self-fullling balance-of-payments crisis, American Economic
Review, Vol. 76, pp. 72-81.
Persons, O.S. (1999), Using nancial information to differentiated failed vs surviving nance
companies in thailand; an implication for emerging economies, Multinational Finance
Journal, Vol. 3 No. 2, pp. 127-45.
Radelet, S. and Sachs, J. (1998), The onset of the East Asian nancial crisis, Working Paper
No. 6680, Cambridge, MA, NBER.
Sachs, J.D., Tornell, A. and Velasco, A. (1996a), Financial crises in emerging markets: the
lessons from 1995, Brooking Papers on Economic Activity, Vol. 1, pp. 147-98.
JRF
8,4
420
Safakli, O. (2003), Basic problems of the banking sector in the TRNC with partial emphasis on
the proactive and reactive strategies applied, Journal of Dogus University, Vol. 4 No. 2,
pp. 217-32.
Saxena, S.C. (2004), The changing nature of currency crises, Journal of Economic Surveys,
Vol. 18 No. 3, pp. 322-35.
Thomson, J.B. (1991), Predicting bank failures in the 1980s, Federal Reserve Bank of Cleveland,
Economic Review, No. 27, pp. 9-20.
Yilmaz, R. (2003) Vol. 20016, Bank runs and deposit insurance in developing countries: the case
of Turkey, PhD thesis, American University, Washington, DC, pp. 1-156.
Corresponding author
Nil Gunsel can be contacted at: ngunsel@neu.edu.tr
The North
Cyprus banking
sector
421
To purchase reprints of this article please e-mail: reprints@emeraldinsight.com
Or visit our web site for further details: www.emeraldinsight.com/reprints
COMMENTARY
Sums of knowledge
Andrew Green
Christofferson Robb & Company, London UK
Innity
Though it is hard to make the following comparison with any accuracy, I will posit that
we live in an age of relative condence: in our achievements, our abilities, our
entitlements, and our political positions basically, in our knowledge. I cannot pretend
to know why, and the proof must be the subject of a different paper. But I believe there
were times, perhaps two or more centuries ago, during which such condence was
weaker. When some of the greatest human insights were about to be made, people
generally knew quite a bit more about what they did not know. Indeed, awareness of
ignorance must have been precisely what drove Renaissance thinkers to overcome it. A
hunger for knowledge, says logic, requires an apparent shortage. Compared to earlier
eras, if I may generalize, ours seems to be dened by how little we care to know about
what we do not know. And if that is the source of our condence, then we should give it
another name: hubris or pigheadedness, perhaps; or plain old stupidity.
Modern society encourages us to believe that the Earth circles the Sun, that medical
science has found cures for ninety percent of what ails us, and that we can predict the
weather hundreds of years into the future. The rest of discovery has been written off as
a simple matter of application. Such condence arises from our perfect understanding
of pretty much everything, and it touches on a divergent and yet important topic: the
politics of condence, where one mans certainty becomes his license to tell everyone
else what to do but that also must wait for another paper. Back on the topic of risk,
let us say that the mathematical expression for such an abundance of (apparent)
knowledge is innity, and the quantity of associated unknowns is zero.
One over innity
The truth, as any amateur philosopher should know, is different. The Earth does not
circle the Sun; instead, both masses wobble and weave through space-time (for now,
that is, until a better physicist than Einstein comes along). As for medical science, even
a fundamental organ such as the liver presents it with a conundrum; hour by hour a
liver will crank out millions of chemicals, few of which can be identied, by processes
as yet undescribed, for purposes unknown. If we cannot say how the body works, how
can we think to cure all its ills? Finally, the weather: thanks to television, we regard it
as a simple mechanism, easily directed by the remote control in the meteorologists
hand; yet it, too, remains a great puzzle. Gasses, cosmic rays, Newtonian mechanics,
quantum physics, and who knows what else, mix and move in ways that climatologists
will never completely understand. When we unmask old mysteries, all we nd
underneath are new ones. And the new ones tend not to be smaller, but bigger.
Of course, there are many things we know for certain; they are the causes and
effects we can see repeating around us, described by our words in such a way that the
cause-effect relationships become logical inevitabilities. Wittgenstein (1922) was the
JRF
8,4
422
The Journal of Risk Finance
Vol. 8 No. 4, 2007
pp. 422-424
qEmerald Group Publishing Limited
1526-5943
DOI 10.1108/15265940710777351
rst to say this: we can only state the obvious, because our words mean what we say
they mean, and the rest is. . . well, we cannot say, because we have no words for it.
There is only the obvious, or nothing knowledge is expanding all the time, but it
remains binary in that sense. The boundary between absolute certainty and total
ignorance exists because humans reside inside the system they are trying to describe,
which means they can never describe it completely. The information age soothes us by
telling us we know it all, pretty much, but we cant possibly and that means, in the
grand scheme, we know essentially nothing. The mathematical expression for our
actual, versus our apparent, knowledge, is a very small number (what we really know)
out of a very large one (what there is to know); essentially, one over innity.
Cynicism and common sense
Let us apply this humbling inventory of knowledge to one of todays (literally) hot
topics: global warming. What should an actuary do when calculating premia and loss
ratios for hurricane cover, in the context of a theory (for want of a better word) about
human action making carbon dioxide, in turn making oceans hotter, in turn powering
tidal surges? Likewise, what should a potential investor in hydrogen fuel cells assume,
not only about hydrogen-for-petroleum substitution, but also about the economics of
the fuel-cell business, including hydrogen manufacturing plants, steel storage tanks,
electrodes, and motors?
Well, the rst motto is cynicism. The actuary should assume the worst not about
events to come, but about the person who is trying to predict those events. What does
the average expert know? Nothing, really (see above). What is his motive? His own
welfare, surely. How objective will his analysis be? Not at all, because one person is a
subjective instrument of thought; a creature that thinks for itself, by every possible
meaning of those words. On top of that, how many steps of insight are required for this
person to get from todays weather to tomorrows? Including the Coriolis effect,
gravity, seasons, ocean currents, salinity, land contours, subatomic-particle
bombardment, cloud formation, light absorption, and the chemical reactions among
the thousands of gases that make up the atmosphere, there are a vast number of
factors all of them necessary just to move time forward by twenty-four hours. Is it
wise, therefore, to accept, for the purposes of selling insurance cover, someones feeble
few calculations, which probably are done badly anyway, and a thermometer dip or
two, and use them to extend time forward (or backward) by thousands of days, or
years? Such a prediction must be worthless; once more, the sum of knowledge will be
small, and the sum of unknowns innite.
Actually, though, there are certain knowns in this weather-prediction scenario.
They are the cynical truths of the situation; namely, those concerning the selsh actors
and their limited knowledge. With those knowns, the actuary should largely disregard
what the climatologist says, and start looking for more reliable sources of information.
Cynicism is a blend of experience and perspective, so the best sources of information
will be broad and wide, cover a long period of time, and essentially be in conict
with one another. The best information must include the pronunciations of
global-warming acolytes and heretics, both. And the nal judgment reconciliation
of the varied and conicting sources can come only from the person whose money is
on the line, because to delegate that critical decision is to delegate the available prot.
Sums of
knowledge
423
So, a cynic will note that global warming is an industry of its own, with a product to
sell and money to make. He knows this because all human action is industry, with
prot as its motive. He will conclude, therefore, that actual salt-water damage claims
will be lower than the global-warming salesmen are saying, and that premia set during
this twenty-rst century panic will have disproportionately low loss ratios and high
margins, other weather factors permitting. The commercial question then becomes
whether to set rates high, with the herd, and take the available prot, or set them low
and take market share, knowing that extraordinary claims are unlikely to materialize
and that margins will stay positive in spite of the hysteria.
The same cynicism will serve our fuel-cell investor. If he is smart, he will discount
predictions of oil-price spikes and dried-up wells, and put a premium on the
self-interest of the various actors in his commercial model. Again, he should assume
the worst about them. People will burn the fuel that is cheapest for the job at hand, and
will not go out of their way to do anyone else a favor. They will not act for the greater
good, but for their own. It will not matter to them whether the vaunted cell is a net
consumer of oil and a belcher of carbon dioxide (which it is, because of all the
petroleum that must be burned and the water that must be electrolyzed to make
hydrogen, and the storage tanks, catalysts, and motors that must be assembled, and all
the fuel that this manufacture requires).
The investment decision should ignore the background noise and build the plant if
the demand is likely to materialize whether by buyers independent decisions or by
government edict in quantities and prices that will yield a return. Ironically, if there
were such a thing as a truly disinterested ecologist, one with nothing to sell or gain
(and remember, there is not), he would see todays fuel-cell technology as a monument
to waste, because it is so dependent on oil compared to the gasoline engine. But a
clear-headed investor might do his sums and see that all is well.
For both actuary and investor, the apparent risks are not the real ones. The
apparent knowledge isnt knowledge, and the experts are so mainly when it comes to
serving themselves. In other words, the commonest of common sense rules. That is the
way of the world, and the best approach to dealing with the risks life poses is to remain
absolutely certain of it.
Reference
Wittgenstein, L. (1921), Logisch-Philosophische Abhandlung, Annalen der Naturphilosophie,
Vol. 14; Tractatus Logico-Philosophicus (translated by C.K. Ogden).
JRF
8,4
424