Retail Credit Scoring: Context and Issues

Literature review

Retail credit scoring: context and issues

1
Retail Credit Scoring: Context and Issues

In financial markets, lending processes are based on actions that involve two

principal parties. According to Franěk and Semerák (2005), these actions range from

the initial loan application, acquisition of the loan to the successful and complete

repayment of the loan or in other cases to its default. Kočenda and Vojtek (2009)

contend that although “retail lending is among the most profitable investments in

lenders' asset portfolios (at least in developed countries), increases in the amount of

loans also bring increases in the number of defaulted loans” (p. 1). This implies that

defaulters have to pay more in terms of interest rates as a way of ensuring the

profitability of the lending institutions. Before a credit is granted, the financial and

lending institutions have to differentiate between “high risk” and “low risk” debtors.

Although asymmetric information between borrower and the lender can cause some

issues in credit lending, parametric or non-parametric credit-scoring methods can be

used.

Credit Scoring Defined, Retail Credit Scoring

Anderson (2007) proposed that to characterize credit scoring, the term ought to

be separated into two segments, credit and scoring. Essentially, the word credit signifies

"purchase now, pay later". It is gotten from the Latin word credo, which implies I

believe or I trust in. Furthermore, the word, scoring alludes to "the utilization of a

numerical instrument to rank request cases as indicated by some genuine or saw quality

to segregate in the middle of them, and guarantee objective and reliable choices". In

this way, scores may be displayed as "numbers" to speak to a solitary quality, or

"evaluations" which may be introduced as "letters" or "names" to speak to one or more

qualities (Anderson, 2007). Hence, credit scoring can be just characterized as " the use

of statistical models to transform relevant data into numerical measures that guide credit

2
Retail Credit Scoring: Context and Issues

decisions” (Abdou & Pointon, 2011, p. 65). Besides, “credit scoring is the use of

statistical models to determine the likelihood that a prospective borrower will default

on a loan. Credit scoring models are widely used to evaluate business, real estate, and

consumer loans" (Gup&Kolari, 2005, p. 508). Likewise, "Credit scoring is the set of

decision models and their underlying techniques that aid lenders in the granting of

consumer credit. These techniques are used to decide who will get credit, how much

credit they should get, and what operational strategies will enhance the profitability of

borrowers to lenders" (Thomas et al., 2002, p. 1). To sum it up, credit scoring is a

method used to compute the credit qualification of a consumer and how much should

be lend.

Credit Scoring according to Lessmann, Seow, Baesens, and Thomas (2013) is

concerned with the assessment of financial risks and using that information to inform

managerial decision making in the money lending business. In general, a credit score

can be described asa model-based estimate of the probability employed to show the

borrower’s behaviour in the future (Lessmann et al., 2013). So as to determine the credit

worthiness of a customer applying for a credit, the ban is compelled to employ a

scorecard with the aim of estimating how likely credit applicant can default (Crook et

al., 2007). Typically, the word“credit scoring” is used in reference to credit decisions

made in the retail business. Kumar and Ravi (2007) points out that management of

national or corporate credit is very important in the financial sector. Thus, there are two

major uses of credit scoring namely: predicting the credit worthiness of a customer, and

the reliability of that customer. Since thousands of customers apply for credits every

day in financial institutions, credit scoring is the most effective which can be used to

determine the reliability of the customer. Řezáč, and Řezáč (2012) explains that “the

assessment of the risk associated with granting of credits has been underpinned by one

3
Retail Credit Scoring: Context and Issues

of the most successful applications of statistics and operations research” (p. 2).

Therefore, credit scoring focuses on credit worthiness of borrowers based on statistical

calculations.

When applied, credit techniques determine persons to get credit and who not to

get, and how much credit should be given to those who qualify. Credit scoring

techniques are used to assess the lending risk (Řezáč, &Řezáč, 2012). In this context, it

is applied to identify “good” or “bad” applications on a personal basis, but they forecast

probability, that an applicant with any given score will be “good” or “bad”. Řezáč, and

Řezáč (2012) have noted that these scores, supported with other business considerations

like losses ad expected approval rates, are employed when making decision making.

The retail business is of extensive significance. For instance, the aggregate

volume of purchaser credits which were held by business banks in the U.S 2013 was

$1,132.4 bn; contrasted with $1,540.6 bn in the corporate business. Conversely, in the

UK, advances and home loans to people were much higher than corporate advances in

2012 (£11,676 m c.f. £10,388 m) (Lessmann et al., 2013). Nielsen (2012) explained

that the credit scoring is likewise utilized to support new charge cards. On a worldwide

scale, the aggregate number of charge cards coursing in 2011 was 2,039.3 m, and these

were utilized as a part of 6424 bn exchanges. Particularly developing nations have seen

amazing development figures in the quantity of Mastercards in the later past (Akkoc,

2012). Given these figures, it is clear that budgetary organizations discriminatingly rely

on upon quantitative choice guides to empower precise danger evaluations in purchaser

giving. This is the extent of credit scoring, which Crook et al. (2007,) consider as “one

of the most successful applications of statistics and OR in industry” (p. 1448)

Credit scoring models in retailing usually play a significant role in the risk

management practice of most banks. For example, they are employed to quantify credit

4
Retail Credit Scoring: Context and Issues

risk at transaction level in an array of phases of the credit cycle (Abdou & Pointon,

2011). Such models qualify to behavioral and collection models. When statistical

calculations are carried out, accredit score is derived which is used to empowers users

in making effectiveand quick decisions or even automating decisions. Kraus (2014)

explains thatthis is desirable when banks are dealing clients and small margin of profits

derived from individual consumer lending. Abdou and Pointon(2011) add that most

financial institutions and banks, across the world, have lately developed or changed

their existing internal credit risk models to adaptto the new rules and best practices

being currently used in thefinancial and banking market.

Bofondi and Lott (2005) have explained that in retail, credit scoring ensure that

costs of underwriting process are reduced, and accuracy in estimating borrower’s

default probability improved. This is achieved after scores have been generated based

on borrower’s characteristics such as previous loans and income. Wang (2010) explains

that “A well-designed model should give high scores to borrowers whose loans would

perform well and low scores to borrowers whose loans would not perform well” (p. 13).

This means when credit scoring is applied in retailing, it provides the companies with

scores which shows low and high scores of the borrowers.

Benefits and criticism of credit scoring

Based on the existing literature, there are a number of benefits which are

associated with credit scoring techniques when employed in retail context. Some of

these benefits are important to both borrowers and lenders.

Wang (2010) explains that credit scoring results to an increased efficiencies and

costs to credit officers. For instance, when giving loans or credit facilities to customers,

credit scoring saves time because it provides a list of the most credit worth customers

and those who are not. Moreover, it is an efficient way to granting loans, as it reduces

5
Retail Credit Scoring: Context and Issues

the time, in terms of days and weeks to hours. This makes retail credit scoring a standard

loan granting process (Abdou & Pointon, 2011).For instance, both good and bad

characteristics of borrowers are considered. Wang (2010) contends that “model helps

lenders ensure that the same underwriting criteria have been applied to all borrowers

regardless of race, gender or other factors prohibited by commercial law used in credit

appraisal” (p. 14). What this means is that retail credit scoring models allow lenders to

give credit facilities based on bad and good characteristics of the borrowers regardless

of their personal information and backgrounds.

Credit scoring requires less data to settle on a choice, in light of the fact that

credit scoring models have been evaluated to incorporate just those variables, which are

measurably and/or fundamentally associated with reimbursement execution; though

judgemental choices, by all appearances, have no factual noteworthiness and in this

manner no variable lessening routines are accessible (Crook, 1996). Credit scoring

models endeavour to amend the predisposition that would come about because of

considering the reimbursement histories of just acknowledged applications and not all

applications. They do this by expecting how dismisses applications would have

performed on the off chance that they had been acknowledged. Judgemental strategies

are normally in light of just the qualities of the individuals who were acknowledged,

and who hence defaulted (Crook, 1996). Credit scoring models consider the attributes

of good and in addition terrible payers, while, judgemental strategies are for the most

part one-sided towards attention to awful payers. Credit scoring models exhibit the

relationship between the variables included and reimbursement conduct, though this

connection can't be shown on account of judgemental systems in light of the fact that a

hefty portion of the attributes which an advance investigator may utilize are not

unbiased measured (Bartolozzi, et al., 2008). A credit scoring model incorporates an

6
Retail Credit Scoring: Context and Issues

expansive number of a customer’s attributes at the same time, including their

associations, while an advance analysts mind can't apparently do this, for the errand is

excessively difficult and complex. An extra vital advantage of credit scoring is that the

same information can be broke down effectively and obviously by distinctive credit

investigators or analysts and give the same weights. This is profoundly unrealistic to be

so on account of judgemental systems (Bartolozzi, et al., 2008).

Some different benefits of credit scoring has been condensed by Al Amari

(2002), referring to Chandler & Coffman (1979) as takes after: more productive

handling time, and consequent backing for the choice making procedure; minimization

of credit procedure expenses and exertion; less slips made; procurement of estimations

to be thought about in post reviews; incorporation of variables upheld through target

investigation to survey the credit danger; demonstrating taking into account genuine

information; interrelation between variables are viewed as; less client data requirements

for credit choices; the cut off score (more subtle elements of cut-off scores are given in

the following subsection) can be changed by elements influencing the keeping money

division; acknowledgment of just approved elements considered by surely understood

organizations, for example, ECOA in the United States and Consumer Credit Act in the

United Kingdom; and the decision of data bolstered as being identified with client credit

hazard.

Giving is portrayed by data asymmetry in the middle of banks and borrowers.

Borrowers have an instructive favourable position over moneylenders in light of the

fact that they know more than banks do about the venture ventures they wish to

embrace. This enlightening point of preference results in unfriendly determination and

the exemplary "bitter" issue (Akerlof, 1970) that happens in light of the fact that banks

can't separate "great" borrowers (low default hazard) from "awful" ones (high default

7
Retail Credit Scoring: Context and Issues

hazard). Banks will accordingly just make credits at premium rates reflecting normal

borrower quality, with top notch borrowers paying a premium rate that is too high and

low quality borrowers paying a premium rate that is too low. Hence, some brilliant

borrowers may choose to drop out of the business sector. Unfavourable choice can

likewise be depicted as an issue that happens on the grounds that potential terrible

borrowers are the ones that most effectively look for advances which builds the chance

that an advance may be made to an awful borrower (Bartolozzi, et al., 2008). Stiglitz

and Weiss (1981) break down unfavourable choice and motivating force impacts in the

advance market and demonstrate that a bank that raises its advantage rate may endure

unfriendly determination in light of the fact that just dangerous borrowers will be

willing to get at higher rates. Subsequently, banks may pick not to raise their advantage

rates to take out credit interest, bringing about conceivable credit apportioning

Amid a monetary and money related emergency, the unfriendly choice issue

commonly increases on two fronts. In the first place, the level of data asymmetry

increments as banks think that its more hard to recognize great from terrible borrowers.

The conduct and execution of borrowers are required to change fundamentally amid an

emergency; however banks experience issues in effectively reckoning these

progressions and fusing them into their credit choice making procedure. Second,

expanded default danger interprets into a higher danger premium. These higher

acquiring expenses thusly cause all the more great borrowers to drop out of the business

sector (Bartolozzi, et al., 2008).

Research Gap and Formulation of a Specific Research Question

Abdou and Pointon (2011) in their study have pointed out that the literature of

credit scoring is very limited, and its history very short. This is because the number of

8
Retail Credit Scoring: Context and Issues

books discussing credit scoring and the international journal articles that have explored

different credit techniques are few.

Most of the credit-scoring literature according to Kočenda and Vojtek (2009) is

related to non-retail loans. This is because data lined to loans provided to firms is more

available. Moreover, corporate credit scoring literature is more common compared to

retail-credit retailing because scoring for retail loans for several reasons. Primarily, the

amounts lent are much smaller in the case of retail lending, and therefore from the point

of view of risk management, retail loans are dealt with using a portfolio approach, while

corporate loans are managed on an individual basis.

Franěk and Semerák (2005) have observed that “empirical studies on credit

scoring with respect to retail loans are infrequent in the relevant literature on developed

markets and, to our best knowledge, no such empirical study exists with respect to retail

loans in post-transition countries” (p. 2). This implies that limited empirical studies

have been done with regard to credit scoring in retailing, and for this reason, there is

need for such research studies. This is because there is limited data available since

privacy laws provide protection because of sensitivity of information associated banks’

policy to release client’s information. The authors have added that majority of the

credit-scoring literature deals with industry loans. This means that it has focused on

loans received by firms, rather than retail credit scoring. In addition, there is very

minimal research which has been done on banks to test retail credit scoring.

Based on the research gaps existing in the current literature review, the

proposed research question is: In what ways can retail credit scoring be applied in the

naming sector to determine loan defaulters and credit worthy customers?

9
Retail Credit Scoring: Context and Issues

Rationale and Importance of the Research Question

By answering the research question we will gain information related to retail

credit scoring and how banks use it to determine the credit worthiness of their clients.

In addition, answering this question would provide us with information on the primary

characteristics of customers which are employed when determining customers who

qualify for credit facilities. At the end of the study, it would be possible to show the

difference between bad credit and good credit. Moreover, it would be possible to

determine why retail individuals’ borrowers do not get the same benefit of borrowing

compared to corporate defaulters. Successful answering of this question will provide to

the existing credit scoring literature because there is very limited studies which have

tried to answer that question because of the sensitive nature of information and data

relating to customers.

Retail credit scoring is based on statistical methods used for analytical purposes.

Such methods include prohibit regression, logistic regression, and discriminant

analysis. These methods are to be used in this study because two datasets have are to

be used. In addition, the sample will be divided based on observed characteristics such

as level of income, family size, and age among others.

Research Methodology to be used to address the Research Question

In general credit scoring entails the generation of number points (scores) based

on the number of borrowers relevant characteristics including previous loans, wealth,

age, profession, and income. Wang (2010) points out that “Credit score is based on

statistical analysis of the borrowers’ credit files, using borrowers’ historical data and

statistical techniques; it tries to isolate the effect of various characteristics of applicators

on delinquencies and defaults” (p. 1). This means that for a study in retail credit scoring

to be successful, statistical data and calculations have to be present. Thus, the study will

10
Retail Credit Scoring: Context and Issues

empirically investigate the problem identified and answer the research question. The

research is to be based upon the quantitative and qualitative approach. As such, both

qualitative data (demographics of the customer) and qualitative data (credit worthiness,

loans, defaulting among other variables) will be employed and tested.

Data Description to be Test the Research Question

Since the question is lined to retail credit scoring, the focus will be in the

application scoring data to determine whether the client should be granted credit or not.

Thus, the necessary data required in this case will be derived from credit information

provided by banks and customers’ information. Customers are more likely to give out

their information because they have interest in credit facilities and products. Thus, the

dataset to be employed in the analysis will be provided by a bank that specializes in the

provision of small and medium sized loans to retail customers. The information can

also be the same that was used by the bank to provide scoring and assessment models.

In this data, various demographic characteristics as well as other information collected

from the targeted customers. In addition the dataset should be between 2007 to 2010 a

period when the credit crunch had hit the United States and Europe very hard. The set

should contain customers who repaid their loans and those who defaulted. This is based

on Thomas et al. (2002) advice on data arrangement where the sample should be spit

50/50 between bad and good loans. The rationale behind the splitting is to keep the

same odds in the sample to be used, which is a representative of the population. Other

variables to be included to support personal information include the customers’ debt

position, age, marital status, income, payment history.

11
Retail Credit Scoring: Context and Issues

Reference List

Abdou, H. &Pointon, J. (2011) 'Credit scoring, statistical techniques and evaluation

criteria: a review of the literature ', Intelligent Systems in Accounting,

Finance & Management, 18 (2-3), pp. 59-88.

Akerlof, A (1970). The Market for "Lemons": Quality Uncertainty and the Market

Mechanism. The Quarterly Journal of Economics, Vol. 84, no. 3, pp. 488-

500.

Akkoc, S (2012). An empirical comparison of conventional techniques, neural

networks and the three stage hybrid Adaptive Neuro Fuzzy Inference System

(ANFIS) model for credit scoring analysis: The case of Turkish credit card

data. European Journal of Operational Research, vol. 222, no. 1, pp. 168-

178

Al-Amari, A. (2002). The credit evaluation process and the role of credit scoring: A

case study of Qatar. Ph.D. Thesis, University College Dublin.

Anderson, R. (2007). Credit Scoring Toolkit: Theory and Practice for Retail Credit

Risk Management and Decision Automation. Oxford University Press,

Oxford.

Bartolozzi, E, Cornford, M, Garc´ıa-Erg¨u´ın, L, Deoc´on, P, Vasquez, O I, & Plaza,

F J (2008).Credit Scoring Modelling for Retail Banking Sector.[Online]

Available at: http://www.mat.ucm.es/momat/2008mw/creditscoring.pdf

(Accessed 3 Aug. 2015).

Bofondi, M & Lott, F (2005).Innovation in the Retail Banking Industry: Credit

Scoring Adoption and Consequences on Credit Availability. Bank of Italy,

Research Department

12
Retail Credit Scoring: Context and Issues

Chandler, G.C., & J.Y. Coffman (1979), ‘A comparative analysis of empirical vs.

judgmental credit evaluation’, Journal of Retail Banking, 1, pp. 15-26.

Crook, J. N. (1996). Credit scoring: an overview., Working paper no. 96/13,

Department of business studies, The University of Edinburgh.

Crook, J.N., Edelman, D.B., & Thomas, L.C. (2007). Recent developments in

consumer credit risk assessment. European Journal of Operational

Research, vol. 183, no. 3, 1447-1465

Franěk, P &Semerák, V (2005). Credit Scoring methods. [Online] Available at

https://www.risknet.de/uploads/tx_bxelibrary/CreditScoring.pdf (Accessed 3

1 2015).

Gup, B. E &Kolari, J. W. (2005).Commercial Banking: The management of risk.

Alabama: John Wiley & Sons, Inc.

Kočenda, E &Vojtek, V (2009). Default Predictors and Credit Scoring Models for

Retail Banking, CESIFO working paper no. 2862 .

Kraus, A. (2014) Machine Learning for Credit Scoring. [Online] Available at:

http://edoc.ub.uni-muenchen.de/17143/1/Kraus_Anne.pdf (Accessed 3 1

2015).

Kumar, P.R., & Ravi, V. (2007). Bankruptcy prediction in banks and firms via

statistical and intelligent techniques: A review. European Journal of

Operational Research, vol. 180, no. 1, pp. 1-28.

Lessmann, S, Seow, H, Baesens, B, & Thomas, L. C (2013) ‘Benchmarking state-of-

the-art classification algorithms for credit scoring: A ten-year update’,

[Online] Available at: http://www.business-

school.ed.ac.uk/waf/crc_archive/2013/42.pdf (Accessed 3 Aug. 2014).

13
Retail Credit Scoring: Context and Issues

Nielsen. (2012). Global Cards — 2011. The Nielsen Report, April 2012 (Issue

992),Carpinteria, CA, USA.

Řezáč,, M &Řezáč, M (2012) How to Measure Quality of Credit Scoring Models

[Online] Available at:

http://www.insightdata.com.au/mirror/whitepapers/Rezac.pdf\ (Accessed 3

Aug. 2014).

Stiglitz , J. E., & Weiss, A (1981). ‘Credit rationing in markets with imperfect

competition’, American Economic Review, vol. 71, pp. 393-410.

Thomas, L. C.,. Edelman D. B & Crook, L N (2002), “Credit scoring and its

applications, Philadelphia: Society for industrial and Applied Mathematics.

Vojtek, M &Kocenda, E. (2005).Credit Scoring Methods [Online] Available at:

https://www.risknet.de/uploads/tx_bxelibrary/CreditScoring.pdf (Accessed 3

Aug. 2015).

Wang, W (2010). The Probability of Chinese Mortgage Loan Default and Credit

Scoring, Thesis, Lincoln University

14