African Journal of Business Management Vol. 7(4), pp.

265-272, 28 January, 2013

Available online at http://www.academicjournals.org/AJBM
DOI: 10.5897/AJBM12.1396
ISSN 1993-8233 ©2013 Academic Journals

Full Length Research Paper

Credit risk management of commercial banks in Iran;

Using logistic model
Hamid Sepehrdoust* and Adel Berjisian
Department of Economics, Bu-Ali-Sina University, Hamedan, Hamedan Province, Iran.
Accepted 19 December, 2012

The observed crisis and profitability decreases in banking system are mainly because of the
inefficiency in credit risk control and that's why, utilization of customers' ranking system is the most
important tool that is required for managing and controlling the risk. The goal of the study was to
present an applied model for credit scoring of real entity customers of banks with reliance on statistical
information of credit customers of Parsian Bank in Iran. For this purpose the logistic regression model
is used to analyze credit ranking and financial scoring of bank’s customers based on their previous and
current data record like; job stability, collaterals, income and some other main indicators for estimating
non-default probability of facilities offered to each customer. The results of the model estimation
showed that non-default probability of facilities have positive relation with variables amount of
collaterals received from customer, monthly income amount of the customer, the status of applicant for
taking facilities such as place of residence (be owner or tenant of the applicant), the age of applicant for
taking facilities, occupational status of the applicant as stability and educational level of the applicant
for taking facilities and have negative relation with amount of paid facilities to the customer and
payback duration of granted facilities to the applicant.

Key words: Risk management, credit risk, commercial bank, logistic regression.

INTRODUCTION

Financial sector is the backbone of economy of a country. the credits need of the economy. Both theoretical and
It works as a facilitator for achieving sustained economic empirical evidence suggests a positive correlation
growth through providing efficient monetary inter- between real economic growth and banks assets-
mediation. A strong financial system promotes invest- especially credits (Alashi, 1991). On the other hand since
ment by financing productive business opportunities, a considerable part of banks‟ income are resulted from
mobilizing savings, efficiently allocating resources and absorbing resources of investors and granting facilities
makes easy the trade of goods and services (Jah and from the credits, therefore, they are always encountering
Xiaofeng, 2012). Banks as institutions that are looking for with an important and challengeable issue of credit risk at
maximum profitability have responsibility for increasing the time of granting facilities. Many studies show that,
the value of shareholders‟ equities on one side and evaluation of banking system customers risk in Iranian
attracting the customer satisfaction in the other side. The banks relies on expert's judgment and fingertip rule. This
role of the banking industry is crucial in the pattern and type of evaluation resulted in high rate of postponed
pace of economic growth and development (Mogboyin et claims (Salehi and Mansoury, 2011).
al., 2012). As it is well articulated and established in the Risk is a concept that denotes a potential negative
literature (Demirguc-Kunt and Huizinga, 1997), banks impact to some characteristic of value that may arise
occupy a position in the financial system that supplies from a future event (Mojtahedi et al., 2009), or one can
say that exposure to the consequences of uncertainty
constitutes a risk. In everyday usage, risk is often used
synonymously with the probability of a known loss. Risk
*Corresponding author. E-mail: hamidbasu1340@gmail.com. communication and risk perception are essential factors
266 Afr. J. Bus. Manage.

foe all human decision making (Cooper et al., 2005). The performance and reduction of credit risk toward
meaning of credit risk is non-receiving cash flows of responding increase in demand for credit products of
granting facilities by banks that exact timely evaluation bank. Credit scoring has obvious benefits in relation to
and examination thereof, especially for average and decreasing the required time for approving facilities that
small facilities and this requires a systematic method. causes increase in its usage in evaluating facilities. This
Credit risk management from the primary stages of method decreases the required time in facilities
granting facilities up to the examination time of approving process. Commercial banking union found that
application for receiving facilities are performed by credit traditional process on facilities approving on the average
experts and also supervision process over granting is 12.5 h for each small commercial facility (Allen, 1995),
credit. In contemporary organizations, risk management while in past the facility takers spend more than two
has become established as a particular method and weeks in this regard. Credit scoring can decrease this
practice of viewing and attending to such risks. As a time to less than one hour. Therefore, Barnet Bank in its
social and cultural phenomenon, in itself, it assumes and annual report confessed that the time spending on
prescribes particular views of the nature, forms, degrees approving facilities for real entity customers or small
and methods of dealing with risky businesses (Giddens, commercial firms that was three or four weeks in the
1999; Power, 2008). In fact, it is required that banks past, could be decreased to less than three hours by
examine the applicant‟s ability in payback of obligations using the above-mentioned system (Lawson, 1995).
and estimating the probability amount of non-fulfillment of Leonard (1995) in his study on a Canadian Bank found
obligations in future (non-default of facilities paid) before that after using this system for 18 months, facilities
any payments to them through credit risk management, approving time that was nine days before stating the use
customers‟ ranking and scoring instruments toward of credit scoring system decreases to three days. Of
decreasing credit facilities risk. According to Basel course, scoring is vastly using in mortgage issues since
committee, credit risk is most simply defined as the the Federal National Mortgage Company encourages
potential that causes a borrower or counter party not to mortgage facilities grantors to use credit scoring system
meet its obligations in accordance with the agreed terms (Dezube, 1996). The goal of this study is presenting an
(Gumparthi et al., 2011). The reason behind it is that applied model for credit scoring of real entity customers
when the facilities are not recovered, the loss resulted of banks with reliance on statistical information of credit
from non-default created and therefore, bank must customers of Parsian Bank.
consider the proper coverage due to loss existence
probability.
Credit scoring is an instrument used to manage the risk LITERATURE REVIEW
that takes action toward ranking of customers by using
quantitative statistics and information of facilities The economic literature of 1950s, defines the word risk
applicants and also statistical techniques (Mester, 1997). and non-assurance as knowledge related to occurrence
With the right credit scoring model, the bank can evaluate and/or non-occurrence of event. From years relevant to
any new or existing profiles of customers accurately, 1980s henceforth, risk and non-assurance get separated
enabling them to minimize potential risks that might be and risk is applied to the condition that there was more
looming. Credit rating systems is used to categorize the than one event for each decision making and the
risk‟s worthiness of a person as high, medium or low. occurrence probability for each event not to be distinct
This allows for decision support by accepting, extending and definite (Greuning and Brajovice, 2003). Risk in
or rejecting any credit request (Sheng and Ying, 2011). banking defines as fluctuation or standard deviation of
Credit scoring models divides the credit applicants into cash flows of a bank and its goal is augmenting
two: good and bad credit groups. Good credit group is a shareholders' equities through acquiring ability toward
group that payback their dues timely and bad credit group achieving commercial and financial goals and maximizing
is a group that will probably not pay their dues (Lee et al., the output after considering the risk. A systematic
2002). The Basel committee has defined credit rating as process of risk management is divided into risk
a „summary indicator‟ of the risk inherent in individual identification, risk analysis and risk response (Li and Liao,
credit, embodying an assessment of the risk of loss due 2007; Duijne et al., 2008). Risk identification requires
to the default of a counter party by considering relevant recognizing and documenting the associated risk. Risk
quantitative and qualitative information. Credit rating, analysis examines each identified risk issue, refines the
through the use of symbols, can be defined as an description of the risk and assesses the associated
expression of the opinion about credit quality of the issuer impact. Finally, risk response identifies, evaluates,
of securities with reference to a particular instrument. selects and implements strategies in order to reduce the
Rating is a measure of credit risk and is only one element likelihood of occurrence or impact of risk events
in investment decision making (Gumparthi et al., 2011). (Mojtahedi et al., 2009). In banking industry, risk is
Some of the benefits of credit scoring system are classified into four main groups including operational risk,
shortening the granting facilities process, more quick commercial risk, event risk and financial risk. Financial
 Sepehrdoust and Berjisian 267

risks are divided into two different risk groups: first group credit scoring method as the first evaluation system of
includes risks relevant to fluctuation in interest rate, credit demand. Dunham (1938) engaged in his studies for
currency and market rate, and second group includes designing a credit risk system, used five important
pure risks as liquidity risk and credit risk and in the case standards including existing condition, revenue condi-
of mismanagement, these two groups directly causes tions, financial conditions, guarantors or Collaterals and
bank's loss (Joel, 2009). facilities payback information of other banks. Considering
Credit risk means a risk resulted from inability of facility these factors, the objective of credit scoring models is to
receiver in payment of the obligations to bank and/or risk assign credit applicants to either a “good credit” group
of non-returning of original and profit amount of that is likely to repay financial obligation or a “bad credit”
investment which caused decrease in current value of group with high possibility of defaulting. Therefore, credit
bank‟s assets (Altman, 1998). Basel Committee working scoring problems are basically a classification problem
under the supervision of international settlement bank of (Johnson and Wichern, 2002). Accurate credit quality
Swiss for assimilating banking rules, defines credit risk as estimation systems will substantially improve the profit-
potential probability in which facility receiver get in- ability of the banking institutions (Thomas et al., 2002).
capable toward fulfillment of its obligations against bank Many studies have been conducted in this regard as
within certain duration (BCBS, 2000).The origin of follows. Durand (1941) in his studies examined which
creation the credit risk may be observed in compilation of variable has been important from facility providers‟ point
three risks that are respectively include: default risk, of view and which specifications are statistically con-
recover risk and exposure risk (Joel, 2009). Default risk siderable for credit risk management. The most important
or non-default probability of debts by the loaner is a loss variables examined included: applicant‟s job, job stability,
if occurred, threatens the bank. Therefore, credit risk residing years at the current place, bank accounts, saving
rooted in probability in default or non-default of facilities and life insurance policies, sex and monthly installments
by facility receiver and its occurrence probability fluctuate amount that the applicant must be paid. Most experts
in the range of zero and one. Payment default informed know Durand as founder of current credit scoring system.
by a bank institution when scheduled installments not to Isaac – Fair Institute (1996) found the necessity for
be paid within a certain duration after due date. Default application and development of credit scoring systems of
may be economic and occurred when economic value of credit facilities‟ customers that needs continuous
assets or the current value of expected future cash flows collecting and updating of data. Bogess (1967) for the
become less than the value of non-deposited debts. Loss first time in an article suggested using computer for
resulted from default is pending on default definition and developing scoring models, evaluation of mass facilities
default definition is pending on estimating default products and increasing algorithm-writing skills which
probability (resulted from past data). Ranking agencies caused possibility for examination and studying a bog
consider the default event after passing three months data collection from different point of views. He examined
from due date of a scheduled payment and no payment complicated tools of multiple criteria statistical methods
performs during this period, therefore, theoretical models which lead to creating much clearer models. Jah and
of credit risk which propounded after Merton model Xiaofeng (2012), compared the financial performance of
(Merton, 1974), apply economic default definition for different ownership structured commercial banks in Nepal
measuring of losses amount. It is noteworthy that, based on their financial characteristics and identify the
different default events necessarily not to create determinants of performance exposed by the financial
immediate loss but increase the probability of permanent ratios for the period 2005 to 2010. The results show that
default or bankruptcy. Default risk measures through the public sector banks are significantly less efficient than
probability of occurring default within certain duration. Of their counterpart are; however domestic private banks
course, default probability may not be measured directly, are equally efficient to foreign-owned (joint venture)
but must be used from statistics collected from default in banks. Owojori et al. (2011), managed to provide an
the past that credited from system interior. overview of risk management practices in insured banks
Usually, previous statistics of default and ranking of in Nigeria. The employed trend analysis of variables to
agencies are proper and accepted standards of default derive its results and concluded by pointing to some
which uses as a symbol of default risk. In the case of steps that would help to preserve the banking system and
non-accessing of agencies‟ ranking, we can use sustain its impact on our fragile economy. Sheng and
estimation about default probability based on specifi- Ying (2011) studied the use of batch and incremental
cations of some applicants (real or legal). Followed by classifiers such as logistic regression, neural networks
John Murray in 1909 and ranking of credit risk on bonds, and C5 to obtain the best classifier to be used for
some researchers realized approximation between bonds improving the predictive accuracy of consumer's credit
and paid facilities and examined the measurement of card risk of a bank in Malaysia.
non-default risk of original and interest of facilities (Kiss, Results showed that C5 emerged consistently as the
2003). In this regard one may point out to the Fisher‟s technique that have generated the best models with an
study (Fisher, 1936) focused on the fundamentals of average predictive accuracy as high as 94.68%. Huang
268 Afr. J. Bus. Manage.

and Wu (2011), studied the customer credit quality decision making makes use of a data bank of previous records of
assessment for banking industries, by using boosting and customer (Dunham, 1938). Type of used information in this model is
pending on different conditions. This model used five major
genetic algorithms (GA). The empirical results indicated
standards for evaluating facility granting to the customers. These
that GA substantially improves the performance of include:
underlying classifiers. Considering robustness and
reliability, combining GA with ensemble classifiers is 1- Character: is a standard for recognizing commitment rate of
better than traditional models. Zribi and Boujelbene applicant for payback of credit facilities. Attended items include
what was the credit background of customer and if he has ever
(2011), aimed to examine the determinants of bank credit
been bankrupt? Whether previous creditors have been referred to
risk in ten commercial banks, taking into account both the court?
macroeconomic factors and microeconomic variables that 2- Income capacity: is a standard for estimating income authority of
are likely to influence credit risk. Overall, the results show the applicant. This standard evaluates financial authority of real or
that the main determinants of bank credit risk in Tunisia legal entity and answers to this important question that if the
is: ownership structure, prudential regulation of capital, applicant has authority for payback of installments relevant to the
applied facilities?
profitability and macroeconomic indicators. The empirical 3- Capital: is a standard for analyzing capital and assets of the
results show that the public ownership increases the applicant which indicates customer‟s authority in payback of
bank credit risk. Other authors that have written in the facilities.
field of designing a risk measurement model include 4- Collateral: is a standard for recognizing type of free assets able
Beaver (1966), Altman (1968) and Morgan (1994). to putting collateral including property and bank collaterals and any
The importance of increasing the degree of credit risk other collateral assure the creditor for compensating non-default
loss.
in Iran also has led to studies in this area. Zekavat (2003) 5- Condition: is a standard for more recognizing of external
in a research inspired by Altman model, examined credit environmental conditions which have effect on payback authority of
risk models of customers of Tosea Saderat Bank. credit commitments of creditors. Main axis of external
Mansouri (2003) in his study used regression and neural environmental conditions is job security and/or job stability of the
networks models for evaluating customer‟s credit risk of applicant.
Mellat Bank. Arabmazar and Roueintan (2006), made an Normally, the financial performance of commercial banks and other
attempt to evaluate the credit risk of legal customers of financial institutions has been measured using a combination of
Keshavarzi Bank Iran and examined qualitative and financial ratios analysis, benchmarking, measuring performance
financial information of a 200-member random sample of against budget or a mix of these methodologies (Avkiran, 1995).
companies which received credit facilities from Bank. Depending on the commercial uses of credit scoring, the
Safari et al. (2010) took action toward presenting a model methodology to construct credit scoring models varies from bank to
bank. It may involve firstly, a sample of historical records classified
for credit ranking of legal customers of Tejarat Bank in as "good" and "bad" depending on their repayment performance
Iran. Salehi and Mansoury (2011), have studied Iranian over a given period. Next, data could be obtained from internal or
banking credit risk and formulated an intelligent model by other external sources, namely, from credit bureau reports. Finally,
neural network and logistic regression to evaluate statistical or other quantitative analysis is performed on the data to
individual customers' credit risk without prejudice and derive a credit scoring model (Koh et al., 2006). In this study the
logistic regression model is used to estimate the model. Logistic
discrimination. The result revealed that neural network
regression is a type of regression analysis used for predicting the
and logistic regression have the same ability in predicting outcome of a categorical criterion variable based on one or more
customer credit risk. The main objective of the present predictor variables. Referred to the target group, the criterion is
study is to present a comprehensive model for credit coded as "0" to a "non-case" and "1" to a "case" in binary logistic
scoring of real entity customers of banks with reliance on regression as it leads to the most straightforward interpretation
statistical information of credit customers of Parsian Bank (Lemeshow and Hosmer, 2000). However, estimation of coefficients
in this method is similar to ordinary regression model, but
in Iran. estimation method thereof is completely different. Because of
nonlinear nature of logistic regression, coefficients of logistic model
evaluate through general maximum likelihood (MLE) method. In
MATERIALS AND METHODS logistic regression, dependent variables are a two-mode variable of
(0 and 1) which allocated itself the amount of zero and one. If we
Data required for credit scoring of real customers have been suppose that Y is a random variable that can give up amounts of
collected from statistical information of Parsian Bank in Iran during zero and one, therefore, probability of Y may be considered
the period of 2008 to 2010. Informative data are usually provided equations (2) and (3).
from data relevant to good pay and poor pay applicants. For
example; in a similar study and through 5-year data of small
commercial facilities less than 5 million dollar, a sample consist of
data relevant to more than 5000 facility applicants related to 17
USA banks for creating a scoring model were examined (Asch,
1995) and variables used in this study are derived from five C
(Character, Capacity, Capital, Collateral and Condition) model and
based on five major standards for evaluating and granting regulated
facilities. According to the five C model, facility providers for The symbol (β‫ )׳‬stands for vector of coefficients and the symbol(x)
evaluating facility granting to its credit customers, for scoring and is column vector of independent variables. The above motioned
 Sepehrdoust and Berjisian 269

equations may be considered equation (4). against granting the facilities such as title deed, cheque, draft,
participation bonds and long-term deposit certificate. Toward
quantifying the above-mentioned variable and using thereof in the
 P 
 ' X
model of collaterals respectively improved liquidity and being
ln  
ensure by encoded and used as virtual variables. Participation

 1  P  (4)
bonds and long-term deposit certificate with code No. 2, title deed
with code No. 1 and cheque and draft with code No. 0 were
defined.
(I): indicates the monthly income of facility applicant that is entered
Equation (4) is an indicator of linear relation between independent
directly into the model.
variables and neoprene logarithm chance ratio. Whereas, chance
(O): indicates the ownership status of the customer regarding his
ratio and logarithm thereof is not calculated directly, therefore,
/her residence place as owner or tenant. In order to quantify this
mentioned coefficient is estimated by maximum likelihood ratio.
variable, the code No. 0 is assigned to the applicant being a tenant
Considering each observation as a Bernoulli probability distribution,
and code No. 1 will be assigned to the applicant being an owner in
the equation (5) must be established for any observation.
the model.
(S): shows job stability of the applicant for facilities or duration that
P(Y  yi )  Pi Yi (1  Pi )1 yi the customer employed at his/her current job, and in the case of
employment at the current job for more than five years, his/her

( yi  0,1)
status from job point of view consider as stable. In this model, if the
customer has stable status with code No. 1 and otherwise, with
(5) code No. 0 was identified.
(T): shows payback duration of facilities or duration within which the
The variable Pi is event probability in ith observing and yi is amount customer must pay back the facilities through predetermined
of random variable which can be zero and/or one (one for installments. This model insert at the model monthly.
occurrence and zero for non-occurrence of event), while supposing (AG): is the age of facility applicant, inserts directly in the model and
that (n) is independent observation, therefore, likelihood equation (DE): indicates the educational degree of the applicant which
will be according to equation (6) and with substituting Pi relevant to toward quantifying and using them as virtual, secondary school
equation (2) in equation (6), we can reach equation (7). diploma and lower code: 0, associate‟s degree and bachelor‟s
degree code: one and master‟s degree and higher code: two, were
allocated.

A comprehensive model is constructed after attaching the weights.

Each client will be rated on each of the parameters based on the
scorecard provided. Each score of the client will be multiplied by the
corresponding weights and a weighted score will be calculated for
each parameter. The weighted scores of all the parameters will be
summed to arrive at the final score of each client. Based on the
final score, the client is given a rating by referring to the rating scale
By taking the natural logarithm of equation (7), we will have: of the model. This final score decides the risk involved in operating
with each client (Gumparthi et al., 2011).
n e ' X n 1 In logistic regression estimation model, independent variable have
ln L  l   yi ln ( 
)   (1  yi ) ln ( ) binary status, it is required to use from minimum sample of 1000-
i 1 1 e ' X i 1 1  e ' X (8)
observation (Whitehead, 2004). Therefore, for estimating suggested
model of credit scoring, 1500 selected files of real entity customers
of Parsian Bank which received credit facilities from bank were
Therefore, estimation of independent coefficients (βi vector) is used. Parsian Bank is the biggest private bank of Iran with more
achieved through maximizing the above-mentioned equation which than 4000000 customers and more than 10 years record has the
is calculated through derivation against each independent variable most granted resources and facilities among private banks. It is
and when each of these derivatives is fixed, it is equal to zero. Of noteworthy that, in the examined level and performing of this
course, the above-mentioned equations have no analytical answer research, Parsian Bank has 279 branches. Due to the fact that
and solving the system of equations through Newton-Raphson default files need to be recognized after the period (date) of
method is possible (Ypma, 1995). Finally, on the basis of presented granting of facilities, and with reference to 150 older branches of
theoretical basics, research model is defined as description of Parsian Bank in Tehran which have more considerable credit files
equation (9). and based on cluster sampling, ten files from each branch including
five files relevant to good pay customers and five file relevant to
Y  0  1 AM  2CO  3I  4O  5S  6T  7 DE  8 AG poor pay customers, whose facilities have been defaulted, were
(9)
selected randomly.
Where;
RESULTS
(Y): is a binary variable with the values of zero and one that indicate
default and nonpayment of facilities. In this study, in order to assess the designed optimum
(AM): is the approved amount of facilities requested by the
customer and examination of customer‟s conditions from character,
model, at the first, the relevant data of the above-
collateral, income capacity, capital and external conditions in mentioned variables for 1500 real entity customers of
facilities committee for payment to the customer. Parsian Bank were interred into the model through trial
(CO): is the amount of collateral which is given to the customer and error method by Eviews7 software. Then, in the
270 Afr. J. Bus. Manage.

Table 1. Goodness of fit criterions.

McFadden R-squared 0.447687 Mean dependent var 0.500000

S.D. dependent var. 0.500167 S.E. of regression 0.346677
Akaike info criterion 0.777668 Sum squared resid. 179.1957
Schwarz criterion 0.809548 Log likelihood -574.2513
Hannan-Quinn criter. 0.789545 Restr. log likelihood -1039.721
LR statistic 930.9389 Avg. log likelihood -0.382834
Prob (LR statistic) 0.000000
* Obs. with Dep.=0 750 Total observations 1500
** Obs. with Dep.=1 750
*Observations with zero value of dependant variable **Observations with unit value of dependant variable.

Table 2. Estimation results of logistic regression model. In addition, Lemeshow-Hosmer test is used to examine
the best fit of the model (Whitehead, 2004). Statistics of
Parameter Estimated value P-value this test examines fit goodness of the model through
C -4/182198 0/0000 grouping. In this examination, all observations of the
AM -0/001231* 0/0000 sample were divided into 15 equal groups (100
CO 3/803806 0/0000 observations each group). Statistics value of this test has
I 0/011367 0/0000 chi-square distribution with k=8 degree of freedom and
O 0/698442 0/0032 achieved in 19.9019 and its probability is 0.0977.
S 0/307626 0/0479 Therefore, the null hypothesis based on equation (10)
T 0 0/0000
was approved and concluded that examined variables
have much good proponent authority.
DE 0/731247 0/0000
AG 0/021451 0/0010
e X '
*Negative sign with the values indicates an indirect relation ( H0 : E y  )
between the relevant variables. 1  e X ' (10)

Therefore and finally the general figure of logistic

estimated model, meaningfulness of coefficients through equation is estimated as equation (11).
Wald statistics, meaningfulness of whole regression
through LR statistics in confidence level of 95% and  p 
Y  ln   4 / 182198 0 / 001231AM  3 / 803806CO  0 / 011367I
nonexistence of co-linearity between the variables and 1 p 
nonexistence of specified errors in the model were  0 / 698442O  0 / 307626S  0 / 012471T  0 / 731247DE  0 / 021451AG (11)
examined and result thereof are inserted in Table 1.
The test of LR statistics is similar to F statistics in linear In this study, Logistic model has been proposed for credit
regression model having chi-square distribution with k=8 scoring of Bank‟s individual customers in order to
degree of freedom (k is the number of independent approve or reject their requests for credit facilities.
variables of model) and is calculated by using Considering the achieved coefficients from applying
formula   2(l  l ) and achieved in 930.99389 as logistic regression model and estimating the equation as
indicated in Table 1. The probability of LR statistics that is ˆ
general form (11), that is required to obtain Y i amount
valued less than 0.05 and near zero indicates that within while putting relevant data of each facility applicants
the confidence level of 0.95, the null hypothesis (Ho) is inside the introduced model and then put this obtained
rejected, that is the regression result is meaningful. Mc. quantity in equation (12) in order to calculate value
Fadden R-square statistics that is similar to R2 statistics
in linear regression derived equal to 0.447687 as shown P̂i
indicators of for the mentioned customer.
in Table 1, which is acceptable considering similar
studies for logistic model. Wald statistics is used for pˆ
evaluating the meaningfulness of logistic regression Yˆi  ln ( i )
variables coefficients. As considered in third column of 1  pˆ i (12)
Table 2, meaningfulness level of Wald statistics for all
achieved coefficients is less than 0.05 which means that
considering zero for all above-mentioned coefficients
P̂
Indicators amount of i with a range of values from zero
were rejected. Therefore, the above-mentioned co- to one (0 - 1) for each facility applicants will be demon-
efficients are meaningful. strative of non-default probability of facilities by the
 Sepehrdoust and Berjisian 271

Table 3. Ranking sample of customers using Credit scoring model.

Rank Customer code Non-default probability Rank Customer code Non-default probability
1 700 0.999999994 1491 518 0.0000001122
2 297 0.999999990 1492 176 0.0000001028
3 1167 0.999999970 1493 24 0.0000000849
4 1165 0.999999960 1494 726 0.0000000657
5 70 0.999999900 1495 344 0.0000000567
6 389 0.999999850 1496 411 0.0000000194
7 293 0.999999800 1497 406 0.0000000048
8 1164 0.999999690 1498 408 0.0000000035
9 1038 0.999999680 1499 1476 0.0000000028
10 783 0.999999620 1500 407 0.0000000010

customer. Evidently, through determining threshold limi- that, credit risk management in banks and financial
tations, if achieved probability is being more than the institutions propound in two levels of dealing with
above determined limit, the possibility of bank facilities individual customers and also portfolio level in which
non-default is considered positive, otherwise, default of credit scoring system is a plan to measure and control
customer‟s facility is recognized to be possible. Of the risk, while dealing at customer‟s level. Performing this
course, such probabilities achieved by the model can be examination and results thereof has importance for a
classified into two types of classification expenses. First bank from different aspects such as shortening facility
type error: when a bad customer is placed wrongly at granting process, more quick performance, and decrease
good customer‟s group and the second type error: when default risk of facilities and most important is acquiring
a good customer is placed wrongly at bad customer‟s indicators for credit risk measurement even individually or
group. portfolio.
As a result of the study, 1500-member of real entity Nowadays, commercial banks in Iran are interested in
sample customers of Parsian bank for facility granting, execution of credit scoring system due to confronting
were ranked as shown in Table 3, using P̂i indicator. For more applicants of facilities and time-consuming of exact
example; in Table 3, ten superior customers with highest evaluation of applicant‟s status and preserving custo-
non-default probability and last 10 customers with highest mers‟ satisfaction. On the basis of model estimation and
default probability were indexed. After calculating the acquired meaningful coefficients, the study comes to the
non-default probability of customer‟s facilities and finally conclusion that, safer received collateral, higher monthly
ranking them, it was observed that 700th customer while income, more stable ownership and occupational status,
obtaining non-default probability in 0.999999994 has higher educational degree and older age of the client on
higher priority against other applicants of the facility. In one hand and lower credit payback duration and amount
addition, non-default probability of 407
th
customer of the value of the facilities on the other hand would
achieved equal to 0.0000000010 and has lowest rank eventually lead to increase in the non-default probability
among applicants for facilities. of credit facilities. That means, non-default probability of
facilities have positive relation with variables amount of
collaterals received from customer, monthly income
Conclusions amount of the customer, the status of applicant for taking
facilities such as place of residence (be owner or tenant
The results of the study indicate that for a sustainable of the applicant), the age of applicant for taking facilities,
growth in Iran, the banking sector of the country has to be occupational status of the applicant as stability and
effective and efficient to respond favorably to the needs educational level of the applicant for taking facilities and
of the productive sectors of the economy. Commercial have negative relation with amount of paid facilities to the
banks as institutions aiming at maximum profitability are customer and payback duration of granted facilities to the
required to examine credit status of customers before any applicant.
payment to the applicants for the purpose of control,
decrease of credit risk and increase in efficiency level of
facility granting process. On the other hand, risks are ACKNOWLEDGEMENTS
events or conditions that may occur and has a harmful
effect, which requires to be effectively adopted for Authors want to thank and appreciate comprehensive
minimizing its undesirable results. supports of Bu Ali Sina University and Parsian Bank
On the basis of discussions made, it has become clear towards their cooperation in performing this research.
272 Afr. J. Bus. Manage.

REFERENCES Koh HC, Tan WC, Goh CP (2006). A two-step method to construct
credit scoring models with data mining techniques. Int. J. Bus. Inform.
Alashi SO (1991). The implications of current monetary policies on safe 1(1):966-118.
and sound banking practice to ensure stability in the industry. NDIC Lawson JC (1995). Knowing the score. US., Banker Mag. November,
Q. 1:25-35. pp. 61-65.
Allen JC (1995). A promise of approvals in minutes; not hours.J. Am. Lee TS, Chiu CC, Lu CJ, Chen IF (2002). Credit scoring using the
Banker 28:23. hybrid neural discriminate technique. J. Expert Syst. Appl. 23:245-
Altman E, Saunders A (1998). Credit risk measurement: developments 254.
over the Last 20 Years. J. Bank. Financ. 21:1721-1742. Lemeshow DW, Hosmer S (2000). Applied logistic regression (2nd ed.
Altman EL (1968). Financial ratios; discriminate analysis and the ed.). New York: Wiley. ISBN 0-471-35632-8.
prediction of corporate bankruptcy. J. Financ. 23:589-609. Leonard KJ (1995). The development of credit scoring quality measures
Arabmazar A, Roueintan P (2006). Effective factors of customer‟s credit for consumer credit applications. International J. Qual. Reliab.
risk; an empirical study of Iran Keshavarzi Bank. J. Jostarhaye Manag. 95:79-85.
Eghtesadi 6:45. Li Y, Liao X (2007). Decision support for risk analysis on dynamic
Asch L (1995). How the RMA/Fair, Isaac Credit-Scoring model was alliance. Decis. Support Syst. 42:2043-2059.
built. J. Commun. Lending 78:10-16. Mansouri A (2003). Designing and explaining bank facilities allocation
Avkiran NK (1995). Developing an instrument to measure customer mathematical model (classic and neural networks model
service quality in branch banking. Int. J. Banks Mark. 12(6):10-18. approach).Doctorate dissertation. Tarbiat Modares University
Basel Committee on Banking Supervision (2000). Principles for the (Teacher Training University). [in Persian].
management of credit risk. www.bis.org/publ/bcbs75.pdf. Mester LJ (1997). What is the point of credit scoring? Federal Reserve
Beaver WH (1966). Financial ratios and predictors of failure. J. Bank of Philadelphia Business Review.
Empirical Res. Account.: Selected Stud. pp.71-111. Merton RC (1974). On the pricing of corporate debt: the risk structure of
Bogess WB (1967). Screen-test your credit risk. J. Harvard Bus. Rev. interest rates. J. Financ. 29:449-470.
45:21-113. Mogboyin O, Asaolu TO, Ajilore OT (2012). Bank consolidation program
Cooper DF, Grey S, Raymond G, Walker P (2005). Project risk and lending performance in nigerian banking system: an empirical
management guidelines: managing risk in large projects and complex analysis with panel data. Int. J. Appl. Econ. Financ. 6:100-108.
procurements. 1st Edn., John Wiley and Sons Ltd., Chichester, Mojtahedi SM, Mousavi SM, Aminian A (2009). A non-parametric
England ISBN: 0-470-02281-7. statistical approach for analyzing risk factor data in risk management
Demirguc-Kunt A, Huizinga H (1997). Financial structure and bank process. J. Appl. Sci. 9:113-120.
profitability: a cross-country comparison of banks, markets and Morgan DP (1994). Bank credit commitments, credit rationing and
development, Demirguc-Kunt A and Levine R (Eds.). MIT Press, monetary policy. J. Money Credit Bank. 26:87-101.
Cambridge, MA pp.243-262. Owojori A, Akintoye R, Adidu A (2011). The challenge of risk
Dezube D (1996). Mortgage scoring: rules of thumb. Mortgage Bank. management in Nigerian banks in the post consolidation era. J.
56:51-57. Account. Taxation 3(2):23-31.
Duijne FHV, Aken DV, Schouten EG (2008). Considerations in Power M (2008). Organized uncertainty: designing a world of risk
developing complete and quantified methods for risk assessment. management. Oxford University Press, Oxford, UK., ISBN-10:
Saf. Sci. 46:245-254. 0199253943, p.275.
Dunham HL (1938). A simple credit rating for small loans. Bankers Safari S, Shaghaghi ME, Sheikh MJ (2010). Credit risk management of
Monthly 55:332-361. legal entity customers in commercial banks; using data envelopment
Durand D (1941). Risk element in consumer installment lending.Studies analysis.Human Sciences Instructor (Modares Oloume Ensani)
in consumer installment financing. No8, National Bureau of Economic Journal. No. 69:137-164. [in Persian].
Research Inc., New York. Salehi M, Mansoury A (2011). An evaluation of Iranian banking system
Fair Institute (1996). Low to moderate income and high minority area credit risk: Neural network and logistic regression approach. Int. J.
case studies, Fair-Isaac and company inc. Discussion Paper. Volume Phys. Sci. 6(25):6082-6090.
4. Sheng KL, Ying WT (2011). A comparative study of data mining
Fisher RA (1936). The use of multiple measurements in taxonomic techniques in predicting consumers‟ credit card risk in banks. Afr. J.
problem. Ann. Eugen. 7:179-188. Bus. Manag. 5(20):8307-8312.
Giddens A (1999). Risk and responsibility. Modern Law Rev. 62:1-10. Thomas LC, Edelman DB, Crook LN (2002). Credit scoring and its
Gumparthi S, Khatri S, Manickavasagam V (2011). Design and applications. Society for Industrial and Applied Mathematics
development of credit rating model for public sector banks in India: Philadelphia, PA, USA.
Special reference to small and medium enterprises. J. Account. Whitehead J (2004). An introduction to logistic regression. Department
Taxation 3(5):105-124. of economics. East Carolina University. p. 10.
Huang SC, Wu CF (2011). Customer credit quality assessments using Ypma TJ (1995). Historical development of newton-raphson method.
data mining methods for banking industries. Afr. J. Bus. Manag. SIAM Rev. 37:531-551.
5(11):4438-4445. Zekavat SM (2003). Credit risk models of Iran Bank-e-Tosea Saderat
Jah S, Xiaofeng H (2012). A comparison of financial performance of customers. Master‟s Program Thesis. Banking Higher Institute,
commercial banks: A case study of Nepal. Afr. J. Bus. Manag. Tehran. [in Persian].
6(25):7601-7611. Zribi N, Boujelbène Y (2011). The factors influencing bank credit risk:
Joel B (2009). Risk Management in Banking. John Willey and Son's The case of Tunisia. J. Account. Taxation 3(4):70-78.
publications.
Johnson RA, Wichern DW (2002). Applied multivariate statistical
analysis (fifth edition).Prentice-Hall, Upper Saddle River, NJ.
Greuning HV, Brajovice S (2003). Analyzing and managing banking
risk: a framework for assessing corporate governance and financial
risk. 2nd edition. The World Bank Publications.
Kiss F (2003). Credit scoring process from a knowledge management
perspective. Budapest University of Technology and Economics.
www.itm.bme.hu.