Francois Grobler, senior consultant at risk management specialists PIC Solutions, South Africa provides an insight
This article presents a case study that features the use of affordability in applications processing in South Africa. Up to now, credit grantors used some measure of affordability to determine the initial credit limit assignment or the maximum loan instalment that the applicant can afford. We provide a case study to determine whether affordability can be used to determine the applicant's likelihood of defaulting, therefore using affordability in the accept/decline decision.
Traditional limit/maximum instalment assignment
At present, credit grantors use the income of the applicant, together with the application risk to determine the initial credit limit or maximum loan instalment of the applicant. The following matrix provides more detail on what such a limit assignment table will look like.
Referring to the above table, an applicant with a Risk Grade of 3 and a gross monthly income of 2500 would thus receive an initial credit limit of 900. The initial credit limits are thus 'tilted' in the table using the risk and income of the applicant, meaning the lower the risk the higher the initial credit limit and the higher the income the higher the initial credit limit.
The Risk Grade could either be an application score alone, or a combination of the application and credit bureau score. This practice has been applied across many credit portfolios with great success.
Reckless lending rules in micro lending
Over the past few years, the South African government and consumer groups have insisted that credit grantors minimise the debt position of consumers. This issue has firstly been initiated and is being monitored at micro lenders. As regulators to the micro lending industry, the Micro Finance Regulatory Council (MFRC) has a dual mandate in ensuring the sustainability of the industry in providing access to finance for lower income individuals as well as ensuring consumer protection.
Micro lending in South Africa is governed by an Exemption to the Usury Act. The Usury Act establishes a consumer protection framework for money-loans and includes a cap on interest rates. Therefore, any credit grantor that operates outside of the Usury Act has to comply with the rules set by the MFRC, resulting in not only micro lenders being 'governed' by the MFRC, but also furniture retailers, some clothing retailers and banks.
The MFRC has set the following rules in terms of reckless lending (in essence, calculating affordability):
The lender shall, prior to entering into a money lending transaction with a borrower, consider the ability of the borrower to make the required payments in respect of the money lending transaction and still to meet his or her necessary living expenses, having regard at least to the following:
Information on the national loan register;
Information provided by the borrower in the application form
Information otherwise disclosed by the borrower to the lender;
The nature of the loan and the purpose for which it is required; and
The borrower's borrowing behaviour reasonably available or known to the lender.
In easy to understand terms, the lender must assess the applicant's affordability, using income, credit bureau information and additional information (as specified in the application form) and use this measure to determine whether the loan should be granted or not. Hence, if the disposable income of the applicant is too low, the application should be declined.
Inherent to this, is the assumption that affordability predicts the applicant's ability to meets his payments to the lender.
Affordability case study
In order to comply with the affordability lending criteria as stipulated by the MFRC and to ensure that the level of bad exposure is improved through responsible lending, a South African furniture retailer commissioned us to conduct an analytical study on affordability
The analytical process that was followed in this study is similar to an application scorecard development.
The data requirements consist of two parts, an application window and a performance window. Below is a graphical representation of the data used in this analysis:
The application window spanned from February 2002 to December 2002, with a normalised performance window of 12 months for every account opened. Affordability was calculated as at application date, using the following calculation:
Rent/bond amount/instalment (sourced from application form)
+ Monthly vehicle instalment (sourced from application form)
+ Existing instalment(s) at the retailer (sourced from host system)
+ Credit bureau instalments (sourced from CCA and NLR databases)
- Housing subsidy (sourced from application form)
= Total monthly commitments
The next step was to calculate an affordability measurement and determine whether this measure would predict the probability of the account defaulting within 12 months of the account opened date.
The affordability measure was calculated as follows:
Total month commitments as a percentage of gross income (sourced from application form).
The accounts were then grouped according to volume distributions and monitored. Below are the results from one of the retailer's chains (the percentages have been changed for confidentiality purposes, but the relationship between the percentages has been kept intact):
The risk grade was determined based on 'joint odds' using an application and credit bureau score. The table indicates that the risk grade rank orders risk, i.e. the lower the risk grade the higher the bad rate. The affordability ratio does, however, not compliment the application decision. One would expect that the bad rate increases as the affordability ratio increase, but this is not the case. In risk grades 1 to 3, accounts with an affordability ratio of 24% or more have a lower bad rate than the accounts with a lower affordability ratio. The risk grade 4-5 column also indicates that affordability cannot differentiate the between high and low risk accounts. This means that one could not use the affordability ratio to determine which applications one should accept or decline. It does, however, prove that the risk grades should still be used to make the application decision.
It is therefore more prudent to implement some sort of affordability measure, whether it is the above formula or only income, in order to allocate the loan amount/maximum instalments on accepted accounts. This should be done by implementing a loan assignment table as specified in the first section of this article.
Model shortcomings
The affordability calculation does have its shortcomings, which could be reasons for the measure not rank ordering risk.
The shortcomings to such a model, specifically in South Africa are:
Credit bureau scores already incorporates elements of affordability (or ability to pay), which minimises the power of the affordability ratio. Characteristics, such as, amounts owed on active accounts, instalments on active accounts and the number of active accounts held by a consumer, are already included in credit bureau scores in order to determine whether the consumer are able to meet his/her debt commitments.
This credit grantor has already had a policy rule in place, whereby applications were referred to manual underwriters if the instalment on the loan applied for, was more than a specified percentage.
The application scorecard includes characteristics related to affordability, such as debt as a percentage of income and a separate income measure.
Due to the relative immaturity of the National Loans Register, the data quality was questionable and thus not adding value to the affordability topic.
Summary
The expected default rate of applicants depends on three major factors, i.e. stability, willingness to pay and ability to pay. The stability factor is largely a function on application form information that relates to demographic information.
The client's willingness to pay can be sourced from the client's payment behavior on other loans or from existing loans at the same institution. The client's ability to pay is purely a function of affordability and can be sourced from external data repositories or from information provided by the applicant on the application form. These three factors are used in combination with each other and not separately, therefore, affordability alone cannot be used in isolation.
This analytical study has provided us with some insight into the use of affordability within an applications processing environment. The data indicates that affordability alone does not predict the likelihood of the applicant defaulting on his/her account, and due to 'double counting' between the affordability formula and application and credit bureau scores, it is not advisable to implement an affordability model to assist in the accept/decline decision. The way forward is to implement an affordability model after the risk decision has been made and apply the model to accepted accounts only and allocate maximum loan amounts or instalments based on affordability.
© Banker Middle East 2006
