Acquiring and managing loan portfolios- including household mortgages, credit cards, automobile loans and all types of equity lines- are major components of the retail credit business.
Different lenders have different means objectives such as minimize required loan loss reserves, minimize the cost of acquiring new customer, maximize the response rate on acquisitions, minimize the default rate on loans and minimize time from billing to collection in managing their portfolios. But the two fundamental objectives of lenders are to maximize profits and maximize market share. Whereas two key objectives of consumer are to maximize the available credit line and minimize the interest rate.
A typical question to consumer about credit card preferences: “Would you prefer a credit line of $25000 with an interest rate of 15% and a credit line of $10000 with an interest rate of 10%”?
Credit scoring offers credit grantors a far more effective technique of managing risk. The more the discrimination provided by credit scoring system, the more credit grantors can avoid type 1 and type 2 errors. On the one hand, credit grantors will reject more bad accounts that would have been accepted; other hand, they will accept more good accounts that would have otherwise been rejected.
As income increases, credit risk reduces. Consumer income and expenditure information give some indication of their likelihood of repayment. But one should need to find out the relationship between payment history and future credit performance. There are some variables that have statistical relationship with credit performance and some have genuine explanatory relationship with credit performance. We should include both types of variables in credit scoring. We select the subset of universe of variables that provide the greatest separation between the mean scores of good and bad. Furthermore, the variables that discriminate directly or indirectly should not be considered in credit scoring.
Accounts are divided into 3 categories – Goods (creditworthy accounts), Bad (uncreditworthy accounts) and indeterminate (neither good nor bad mainly because they are inactive or they are not bad but insufficient time has elapsed to classify them as good). Indeterminate are normally excluded from the scorecard building process.
Chandler (1977) notes a special problem arises in credit-scoring models is that the model developed based on the number of loans granted instead of population of potential borrowers or through-the-door applicants. Chandler, Coffman, Shinkel , Long, Avery (1977) discussed solutions to this problem – one, use both good and bad accepted applicants together with rejected applicants and other, weight the good and bad applicants in accepted group by an estimate of the likelihood that each would be accepted.
Reject inference is the process of attempting to infer the true creditworthiness status of the rejected applicants. Reject inference can be of interest because of determining the number of good credit risks rejected by the scoring instrument and to improve scoring instrument.
Lenders should continue to invest in the development of more sophisticated scoring techniques to better identify customers.
There are some key terms pertaining to retail credit business:
Equal credit opportunity act – A regulation crafted by US government that prohibits creditors from discriminating against credit applicants on the basis of race, religion, color, sex, marital status, national origin, age or receipt of public assistance benefits. This act was enacted in 1974.
Consumer credit protection act – An act to safeguard the consumer through disclosing all terms and conditions by creditors. This act was enacted in 1968.
Risk based pricing – Banks charging high rates were motivated to accept high risks and vice-versa.
Profitability scoring – Accept or reject applications on the basis of net present value of the account. Accept if NPV is positive; reject if NPV is negative.
Acceptance cost = Bad debt cost + Investment cost + Collection cost
Rejection cost = Lost sales cost
Cost of further information = Cost of acquiring additional information + cost of additional decision