This is the html version of the file http://www.cba.edu.kw/elsakka/171-Al-Osaimy_06.pdf.
G o o g l e automatically generates html versions of documents as we crawl the web.


Google is neither affiliated with the authors of this page nor responsible for its content.
These search terms have been highlighted:  profit  measurement  islamic  banking 

Page 1
J.KAU: Islamic Econ., Vol. 17, No. 1, pp. 3-14 (1425 A.H / 2004 A.D)
3
An Early Warning System
for Islamic Banks Performance
M
AHMOOD
H. A
L
-O
SAIMY
Associate Professor
A
HMED
S. B
AMAKHRAMAH
Professor
Economics Department - Faculty of Economics and Administration
King Abdul-Aziz University - Jeddah - Saudi Arabia
A
BSTRACT
. There is increasing demand for predicting the performane of Islamic banks due
to the vital importance of any problem that may face these banks before it materializes and
negatively affects their performance and their financial status. This will save on the costs of
bad performance or failure to depositors, owners and the economy. Thus, a need arises for
an early warning system which will identify the possible causes of bad performance, detect
potential problem banks, facilitate surveilence of banks as well as reduce its costs and make
possible proper timing of examining problem banks as well as scheduling the remedical
procedures. This research aims at benefiting from the previous research efforts on the
subject to develop a preliminary model for the prediction of the performance level of Islamic
banks (i.e. an early warning system), hoping that this will be a cornerstone for further
development and improvisation, specially as more information and data become available or
accessible. To achieve such objective Discriminant Analysis technique will be utilized,
whereby a Discriminant Function will be designed comprising the significant characteristics
(financial ratios) as explanatory variables and the profitability rate as dependent variable.
Discriminant scores are then extracted and used to distinguish between high performance
and low performance groups of banks, thus forming a possible early warning system for the
prediction of future performance of the observed banks. The prediction power of such a
system is finally tested and conclusions drawn.
1. Problem of the Research
There is increasing demand for predicting the performance of Islamic banks due to
the vital importance of prior information on any problem that may face any Islamic
bank before it materializes. This will save on the costs of banks bad performance or
failure to depositors, owners and the society.
Thus, the rationale for an early waring system comes from the following reasons:
1. Identifying the possible causes of bad performance.
2. Facilitating the surveillance of banks and reducing its costs.
3. Proper timing of examining problem banks and scheduling the remedial
procedures.

Page 2
4
Mahmood H. Al-Osaimy and Ahmed S. Bamakhramah
2. Objectives of the Research
1. Using Discriminant Analysis technique to identify the significant characteristics
(financial ratios) which distinguish high-performance banks from low-
performance ones.
2. Designing a Discriminant Function to classify the performance level of the studied
Islamic banks into High performance and Low performance groups based on the
Discriminant scores compared to the Cut-off Discriminant Criterion.
3. Testing the prediction powers of the above Early Warning System.
4. Drawing conclusions on the prediction reliability of the above system and the
degree of its utilization for future prediction of Islamic Banks performance.
3. Review of Literature
Recently an accumulating research on the prediction of business performance
and/or failures has evolved. Discriminant Analysis is one of the most utilized statistical
techniques for the prediction of the performance of business firms. Originally
developed to classify certain variables into two or more pre-specified groups according
to
the most statistically significant distinguishing characteristics (classifying
variables). The discriminant analysis technique usage is extended to the prediction of
the status of such a variable in the future, based on the results of the discriminant
analysis (discriminant function) several years before, mostly between one and two
years prior to the performance or problem or failure occurance, and the the testing of
the classification power of such a function (Altman et al; 1981).
Various models are followed in the discriminant analysis, each with its advantages and
pitfalls, most prominent among which are the Linear Fisher model and the Logit
model (Altman et al, 1981; Amemiya, 1981; Johnsen and Melicher, 1994; Scott, 1981).
Discriminant analysis, though not the oldest technique for the evaluation and
prediction of business performance, being superseded by the Financial Ratios Analysis,
is more preferred to the latter because it gives a summary index of performance, takes
into consideration the possible interrelationship among the characterizing variables
(independent variables) as they explain the variations in the groupings of the classified
variable (dependent variable) and last, but not least, the discriminant analysis can
include other non-financial (e.g. managerial, social or political) factors that may
affect the behavior of the dependent variable (Altmam et al, 1981; Sinkey, 1975).
Lately, discriminant analysis was also applied to the prediction of the performance
and/or failure of financial institutions, markets and instruments (e.g. commercial
banks and investment companies, bond markets and investment portfolios among
others). Although still undergoing fine-tuning improvements, so far the record of such
studies were generally impressive . This was evident from the favourable scores they
acquired in the statistical testing of their classiification results and predictive powers
(Altman et al, 1981; Haslem and Longbrake, 1971; Sinkey, 1975).

Page 3
An Early Warning System for Islamic Banks Performance
5
Haslem and Longbrake (1971) used discriminant analysis to distinguish low
profitability commercial banks from high profitability banks, members of the Federal
Reserve System in the United States, with 46 financial ratios as explanatory variables
and 78 banks (observations) for each group of profitability.
This rersearch aims at benefiting from such efforts to develop a preliminary model
for the prediction of the performance of Islamic banks (i.e. an early warning system)
hoping that this will be a cornerstone for further improvisations and applications.
4. Discriminant Analysis
Discriminant analysis is a statistcal technique used to classify a sample of
observations into two or more groups based on a linear composite of input variables.
In the two group case, the objective of the discriminant analysis is to create a linear
combination of explanatory (discriminant) variables that maximizes the distance
between the centers of the two populations under consideration using the pooled
within-group covariance matrix. This procedure assumes that the explanatory variables
(or predictors) have a multivariate normal distribution with different means but
common covariance matrix among the classes. This linear combination is called
"discriminant function" and can be written as:
z = Σ b
i
x
i
where z is the value of the discriminant function (score), bi's are the discriminant
co-efficients and xi's are the independent (explanatory) variables used to discriminante
between the two groups .
Classifying observations into one of the two groups is done by computing the
discriminant score for each observation and comparing it to a numerical cut-off value.
If the score is above the cut-off value, we assign the observation to group one, and if it
is below the cut-off value, we assign the observation to the other group. The cut-off
value can be computed as follows:
Cut-off value = ( z
1
+ z
2
) / 2 + ln [ c(1/2) p
2
/ c(2/1) p
1
]
where z
1
and z
2
are the centroids of group one and group two respectively, c(i/j) is
the mis-classification cost and Pi is the priori probabilities.
For the case of assuming an equal mis-classification cost, as has been in our
research, the second part of the cut-off value equation is reduced to the natural
logarethem of the priori probabilities ratio. Moreover, if the sample sizes of the two
groups are equal, the cut-off value will be reduced to the mean of the two groups
centroids (i.e. the first part of the equation only ).
Testing the predicting power of the model is usually done on a holdout sample
which is not used in the estimation of the discriminant co-eficients. The discriminant
score is computed for each observation and compared to the cut-off value computed
from the sample used to construct the discriminant function. The pecentage of hits in
the overall grouping is thereafter calculated to obtain the degree of prediction accuracy.
In the absence of a holdout sample, as is the case in our research due to the smallness
of the size of sample banks, other methods were suggested in the literature most
prominent among them is the Lachenbruch method which will be explained later .
i

Page 4
6
Mahmood H. Al-Osaimy and Ahmed S. Bamakhramah
5. Data Collection and Organization
Around 40 letters were sent to various Islamic Banks requesting data on balance
sheets and income statements during the period 1991-1993. These data was used to
formulate performance classification variables and explanatory discriminant variables
(income and expense ratios) which are used to predict the performance of the sample
banks in the year 1993 (the latest year for which complete data on these variables is
available) using data (financial ratios) for each of the previous years 1991 and 1992.
However, only 29 banks responded fully. Out of these, 26 banks with enough operation
experience were selected in the sample, namely those which operated before1989. Such a
relatively small number restricted the size of the observations (number of banks) in each
group of performance. This had obliged us to compress the number of discriminating
(explanatory) variables and aggregate them into block variables which in our belief would at
most represent the operational factors affecting the performance levels of the studied banks.
Another reason which necessitated aggregating the variables is the unstandardized items in
the balance sheets and income (profit and loss) statements of Islamic banks. Only some
banks reported fully detailed assets, liabilities and income entries. Worse than that, some
banks treated investments belonging to customers (depositors) as off-balance sheet items.
Different reporting methods make disaggregate and standardized classification of data
difficult. Hopefully in the future, better detailed and standardized data may enable
researchers to use more disaggregate variables, thus improving the discriminanating
performance of such variables.
With regard to the relative lag in the time-period of the data, two clarifications are due:
1. The set of data in this research paper serves the main purpose of testing the
reliability and efficiency of the early warning system for Islamic banks and does
not incur any conclusions as to the performance of the specific banks covered in
the data. Of course, when the system is to be actually applied by the relevant
bodies as an early warning system, more updated and comprehensive data is
called for.
2. Data on Islamic banks is unfortunately relatively lagging, even from the relevant
banks, a problem which should be addressed in order to render the early
warning system more efficient and reliable in application.
The available sample of banks is divided into two discrete groups: low performance
(problem) banks and high performance (control) banks. The classication (categorization)
criteria is based on a summaryindex of performance to be explained later.
Typically, in discriminant analysis, discrete (dichotomous) variables constitute the
categorization basis of the dependent variable. Continuous variables, like the one we
are using in this research, pose some problems for discriminant measurement because
of the possible arbitrariness in group segmentation, errors in classification tests and
low prediction efficiency (Eisenbeis, 1977).

Page 5
An Early Warning System for Islamic Banks Performance
7
In our case, a success/failure categorization will be appropriate, assigning the
number zero to failure (or problem) group and the number one to success (or non-
problem) group. However, for Islamic banks, data pertaining to failure occurances is
rare because of the rareness of failure cases in the relatively short history of Islamic
banks and the absence of a central regulatory body which has the ample authority to
decide on failure occurance. Thus in our research, we restricted our prediction to
performance, hoping that in the future more detailed, transparent and standardized
information on performance and failure/success occurances will enable our model to be
extended as to function more efficiently as an early warning system for the prediction
of of failure (or problem) cases in Islamic banks.
6. Specification of Variables
Seven financial ratios were used in this research. These ratios were chosen based
on the following criteria:
1. Past similar research on the subject (see references and review of literature).
2. Statistical convenience and efficiency, particularly the problem of the number
observations versus the number of variables (degrees of freedom) mentioned
above, which obliged us to compress explanatory variables to the minimum
aggregate (block) variables.
3. Maximum possible representation of the main factors affecting the performance
of studied banks, mainly productivity, efficiency, liquidity, risk and leverage.
These include only internal factors pertaining to the direct operational income
and expense activities of the studied banks, thus excluding any external factors
e.g. country-specific political, regulatory and policy factors, which are not
directly controlable or predictable by individual banks. These ratios are as
follows:
X
1
= Total Income/Total Assets, representing productivity of bank resources
(resource utilization or asset turnover), where:
Total income includes all income coming from investments (from Islamic
financing e.g. mudarabah, murabahah, musharakah, ijarah etc., or direct financing),
revenues from foreign exchange dealings, revenues from banking services and other
sources (unspecified) of income. Total Assets include liquid assets (cash and reserves),
short term and long term investements and fixed assets.
X
2
= Investment Income/Total Income, representing the level of contribution of
income coming from investments, as explained above, to total income. This variable
distinguishes those banks relying on investment of funds as the main source of income
from those depending on trade (including foreign exchange) transactions and banking
services. When detailed and standardized data become, hopefully, available in the
future, disaggregating the investment income variable into seperate variables (e.g.
mudarabah, murabahah, musharakah, ijarah) will be fruitful
X
3
= Total Income/General and Administrative Expenses, representing the
operational efficiency of the bank, where:

Page 6
8
Mahmood H. Al-Osaimy and Ahmed S. Bamakhramah
General and Administrative Expenses include all operating expenses, staff
expenses, depreciation and provisions.
X
4
= Provisions for Bad Debts and Investments/Total Assets, representing
financing and/or investment risk. Such provisions influence the available funds for
financing and investment and the types of financing or investment, thus affecting the
level of returns from the utilization of these funds.
X
5
= Cash/Total Deposits, representing the liquidity position in the bank which
influences the amount of funds free for financing and investment (against the bank's
eligible liability) and the level of returns, where: Cash includes cash funds in hand
with the bank, balances with other banks (in local and foreign currencies) and reserves
(statutory and others). Due to differences in reported items of liquidity among the
studied banks, liquidity is limited to the above mentioned common items. Total
Deposits include customers current (demand) and investment deposits.
X
6
= Customers Investment Deposits/Shareholders Equity, representing the
leverage level (debt / equity ratio) in the bank’s investment, where:
Shareholders Equity includes paid up capital, reserves and retained profits.
X
7
= Net Profit Before Zakat and/or Taxes/Total Assets, representing the profit
rate of the bank. Zakat and/or taxes were excluded from the calculation of net profit in
order to neutralize the effect of differences in tax and zakat treatment, application and
reporting among the studied sample bank.
The profit rate as measured above is chosen as an idicator of profitability for the
following reasons:
1. Profitabilty includes the income and expense activities of the bank, since profit
equals income minus expenses. Thus, profitability reflects the main ingredients of cash
flow activities in the bank, eminating from the utilization of the bank's resources.
2. The profit rate (net profit/ assets) as a measure of profitability can be
decomposed into two ratios as follows:
Net profit / Assets = (Net profit/Capital) x (Capital/Assets), which mean that the
profit rate is influenced by net profit/capital as a measure of the rate of return for the
bank's shareholders (owners), and capital/assets as a measure of solvency. Both
measures are significant indicators of the success and survival possibilities of the bank.
3. Dividing net profit by assets serves the extra significant purpose of neutralizing
the effect on performance of differences in size among the sample banks, thus diluting
the possible bias in discriminating low performance banks from high performance ones
due to shere size. The technique of matching (pairing) each group of performance
banks according to size utilizing the absolute value of assets, used in many similar
studies, will not be efficient in our sample banks due to the fact that assets are reported
in different currencies.
The dependent variable i.e. the performance level, is classified into two groups:
low performance group and high performance group. Performance is measured in the
form of a summary index composed of the following four financial ratios:

Page 7
An Early Warning System for Islamic Banks Performance
9
Profitability = Net Profit / Total Assets.
Productivity = Total Income / Total Assets.
Efficiency = Total Income / General and Adminstrative Expens
Leverage = Customers Deposits / Shareholders Equity
Classification of the 26 sample banks between the two performance groups is based
on the ranking of each bank according to each of the above four financial ratios,
summing up the ranking scores of each bank for the four financial ratios and
culculating the average score. Those banks with 14 points or less were classified into
the high performance group, while those scoring above 14 points are classifed into the
low performance group. Twelve banks were thus classified into the high performance
group and fourteen banks were classified into the low performance group
7. Empirical Results
Disriminant analyasis was run using the SPSS Discriminant program on the initial
seven financial ratios defined above (X
1
-X
7
) as explanatory (discriminant) variables,
for the 26 sample of Islamic banks to distinguish high performance banks from low
performance ones. The statistics of these ratios for one and two years prior to the
performance year (1993) are presented in Table 1.
Table 1: Univariate Statistics for the Explanatory Variables
High Performance Group
Low Performance Group
Ratio
Mean
St. Dev.
Mean
St. Dev.
One Year Prior (1992)
x
1
x
2
x
3
x
4
x
5
x
6
x
7
0.125
0.103
5.047
0.047
0.254
11.706
0.024
0.130
0.113
2.696
0.093
0.211
6.059
0.022
0.055
0.045
2.584
0.012
0.285
7.539
0.008
0.028
0.023
0.996
0.013
0.205
6.900
0.009
Two Year Prior (1991)
x
1
x
2
x
3
x
4
x
5
x
6
x
7
0.095
0.081
4.629
0.017
0.266
9.118
0.029
0.058
0.065
2.614
0.028
0.177
7.363
0.020
0.064
0.053
3.124
0.012
0.323
9.003
0.015
0.030
0.026
1.569
0.014
0.264
10.933
0.017
Two seperate runs were conducted for the one year prior to the performanace year
data and the two years prior to the performance year data, using the linear discriminant
anlaysis to classfiy Islamic banks performance. The two years prior model did not pass
the statistical significance test of the unequality of the group means (1), and so was
dropped from the analysis. The one year prior model gave impressive results and
passed the significance test of the mean group unequality at 0.05 level of significance,

Page 8
10
Mahmood H. Al-Osaimy and Ahmed S. Bamakhramah
indicating that the two groups (high and low performance groups) came from two
different pupulations. The one year prior model also passed the statistical test of the
equality of the two dispersion matrices at 0.01 level of significance, allowing the use of
linear classification rule. The classification results for the one year prior to the
performance year is given in Table 2.
Table 2: Classification Results for the One Year Prior to the Performance Year
Performance Group
No. of
Cases
Correct
Classification
%
Mis-classification
%
High
12
10
83.3
2
16.7
Low
14
13
92.9
1
7.1
Total
26
23
88.5
3
11.5
The results indicate that out of 12 high performance banks, the model correctly
classified 10 banks. The classification accuracy for the high performance group of
banks is 83.3%, while the mis-classification rate, the type I error i.e. classifying a high
performance as low performance, is 16.7%. For the low performance group of banks,
out of the 14 banks the model correctly classified 13 banks. The classification accuracy
for the low performance group of banks is 92.9%, while the mis-classification rate, the
type II error i.e. classifiying a low performance as high performance, is only 7.1%. The
overall accuracy of the model is 88.5%, which is comparable to most of the studies that
used discriminant analysis.
The relative importance of the explanatory variables used to discriminate between
high and low performance banks will be determined first by applying univariate
statistics to individual variables. Table 3 presents the relative importance of the
individual variables and their ranks.
Table 3: Relative Contributions and Ranks of the Individual Variables
Univariate F
Wilks’ Lambda
Standardized
Coefficient
Variables
Value
Rank
Value
Rank
Value
Rank
x
1
x
2
x
3
x
4
x
5
x
6
x
7
3.88
3.45
10.14
1.97
0.15
2.63
6.39
3
4
1
6
7
5
2
0.86
0.87
0.70
0.92
0.99
0.90
0.79
3
4
1
6
7
5
2
1.35
-1.51
0.51
0.82
0.11
0.80
1.01
2
1
6
4
7
5
3
The results show that the efficiency variable (x
3
) and the profitablity variable (x
7
)
ranked first and second respectively and were significant at 0.05 level in the univariate
F-test. The productivity variable (x
1
) and the investment contribution variable (x
2
)
ranked third and forth respectively and were significant at 0.01 level in the univariate
F- test. The rest of the variables were significant at 0.25 level.

Page 9
An Early Warning System for Islamic Banks Performance
11
Determining the relative importance of the variables by the univariate F-test on
individual variables has a great appeal in the discriminant analysis literature
(Eisenbeis, 1977, pp. 882-884). This is because variables may show little or
insignificant discriminant contribution based on an individual univariate test, but when
combined with other variables they may show high significance. A backward stepwise
method based on the contributions to the multivariate F-test has been proposed. The
results of the backward stepwise method is presented in Table 4. They show that the
risk variable (x
4
) and the liquidity variable (x
5
) were removed. This indicates that these
two variables have no significant power in distinguishing between high and low
performance groups of Islamic banks i.e. high risk and liquidity are common
characteristics of both groups. The other five variables were included by the procedure
and they were significant in the multivariate F- test at 0.01 level. The efficiency
variable (x
3
) and profitability variable (x
7
) were the most important ones to
discriminate between the high and low performance groups.
Table 4: Variable Ranks as they were included in the Backward Stepwise Method
Variable
x
1
x
2
x
3
x
4
x
5
x
6
x
7
Rank
5
4
1
r
r
3
2
r = removed; Multivariate F-Test = 6.21; significant at 0.01
The classification results based on the five variables selected by the backward
stepwise procedure is presented in Table 5. They show that the type I error was
eliminated, but type II error was increased. The five variables correctly identified the
whole set of the low performance banks but mis-classfied three high performance
banks. The results also show that the removal of variables by the backward stepwise
method does not change the overall classification accuracy of the model (88.5%).
Table 5: Classification Results of Islamic Bank Performance for One Year Prior to
the Performance Year
Performance
Group
No. of
Cases
Correct
Classification
%
Mis- classification
%
High
12
9
75.0
3
25.0
Low
14
14
100
0
0
Overall
26
23
88.5
3
11.5
The discriminant function used to derive the classification results in table 5 is:
z = 23.05 x
1
- 23.41 x
2
+ 0.33 x
3
+ 0.09 x
6
+ 45.66 x
7
where z is the discriminant score for each Islamic bank. The classification rule is
to assign an Islamic bank whose score is above 3.14 (the cut-off point) to the high
preformance group and that whose score is below the cut-off point to the low
performance group.
8. Reliability Test
Applying the above discriminant function to data on the five explanatory variables
for an Isamic bank one year prior to the performance year and computing the Z - score
enables the model to function as a predictor of performance of the bank for the
following year.

Page 10
12
Mahmood H. Al-Osaimy and Ahmed S. Bamakhramah
Testing the predicting accuracy of the model is usually done by using a holdout
sample that has not been used in deriving the discrimnant function. In the absence of a
holdout sample, as in our case, various methods have been proposed in the
discriminant analysis literature (Altman et al, 1981: 153-158 ). The most powerful
test, specially for small samples as in our case, is the Lachenbruch method. This
method uses the orignal sample as a holdout sample. The procedure of this method is
to holdout one observation from the original sample each time and use the remaining
observations to derive a function to classifiy the holdout observation. This procedure is
repeated until all observations are classified and then the classification accuracy of the
holdout sample is computed. Applying this method to our observations produced
results similar to those shown in table 5, with 88.5% classification accuracy and 11.5%
expected overall errors .
9. Conclusions
1. The model used in this research proved to be highly efficient in discriminating
between high and low performance Islamic banks groups in spite of the fact that only
26 sample banks and seven initial aggregate financial ratios ( discriminant variables )
were included in the analysis. Five of these variables turned to be signicant in the
discriminant function utilized to test the classification accuracy and prediction
reliability of the model.
2. Reliability of the model as a predictor or as an ealry warning system of
performance of Islamc banks is expected to improve when more detailed and
standardized data become available, allowing for larger disaggregate number of
explanatory (discriminant) variabes to be included. Bank- specific and/or external
variables i.e. managerial, organizational, market, political variables can also be added.
3. The model can also be utilized as an early warning system for various types of
performance including bankruptcy, insolvency and failure, when data on such types
become available and/or accessible.
Notes:
1. The group centroid can be computed by substituting the mean value of the
predictor variables in the discriminant function. This can be written as:
z
i
= b
1
x
1i
+ b
2
x
2i
+ .... + b
m
x
mi
i= 1, 2
where z
i
is the centroid of group i and Xmi is the mean value of the mth
predictor of group i.
2. Testing the unequality of a group mean is the same as testing the significance of
the discriminant function . The null hypothesis is:
H0 : u
1
= u
2
or equivalently Wilks' Lambda = 1
This hypothsis can be tested by an F-test or chi-square test as follows:
F p, N-m-1 = N-M-1 / m . 1-^ / ^ , or x
2 m
= - [ (n-1) - (m+1/2) ] ln^
The null hypothesis is rejected if the computed F or x
2
is bigger than the table
value at 5% level of significance. The computed F is 4.42 and x
2
is 20.5. These
values are bigger than the table values of 2.66 and 14.07 repectively.which leads
to rejecting the null hypothesis.

Page 11
An Early Warning System for Islamic Banks Performance
13
3. The test for equality of group covariance matrices (dispersion matrices) is done
using the method of Box's m and approximate F (Altman et al, 1981: 45 ). The
null hypothesis is:
H0 : s
1
= s
2
where si is the dispersion matrix of group i
i=1,2.
The computed approximate F is 1.67, which is less than the table value of
This leads to accepting the null hypothesis .
References
Altman, Edaward I., Robert B. Avery, Robert A. Eisenbeis and Joseph F. Sinkey (1981),
Application of Classification Techniques in Business, Banking and Finance, Greenwich,
Jai Press Inc.
Amemiya, Takeshi (1981), Qualitative Response Models: A Survey, Journal of Economic
Literature, December, Vol. 29, pp. 1483 - 1536.
Eisenbeis, Robert A. and Robert B. Avery, (1972), Discriminant Analysis and Classification
Procedures: Theory and Practice, Lexington, Mass: D.C. Heath & Co.
Eisebeis, Robert A., (1977), Pitfalls in the Application of Discriminant Analysis in Business,
Finance and Economics, The Journal of Finance, June, Vol. 32, No. 3, pp 875-900.
Haslem, John A. and William A. Longbrake, (1971), A Discriminant Analysis of Commercial
Bank Profitability, The Quarterly Review of Economics and Business, Autumn, Vol. 11,
No. 3, pp. 39-47.
Johnsen, Thomajean and Ronald W. Melicher, (1994), Predicting Corporate Bankruptcy and
Financial Distress: Information Value Added by Multinomial Logit Models, Journal of
Economics and Business, October, Vol. 46, No. 4, pp. 269-286.
Scott, James, (1981) The Probability of Bankruptcy: A Comparison of Empirical Predictions
and Theoretical Models, Journal of Banking and Finance, 5, pp. 317-344.
Sinkey, Joseph F., (1975), A Multivariate Statistical Analysis of the Characteristics of Problem
Banks, The Journal of Finance, March, Vol. 30, No. 1, pp. 21-36.

Page 12
14
Mahmood H. Al-Osaimy and Ahmed S. Bamakhramah
ﻲﻤﻴﺼﻌﻟﺍ ﻥﺍﺪﲪ ﺩﻮﻤﳏ
ﺔﻣﺮﳐﺎﺑ ﺪﻴﻌﺳ ﺪﲪ
ﳌﺍﺴ
ﺺﻠﺨﺘ
:
.