Partners (Fall 1996)


By Robert Kennedy

The phrase "credit scoring" has provoked substantial press in recent months. Credit scoring, also referred to as "automated underwriting", entered the general vocabulary last Christmas when Governor Lawrence Lindsey of the Federal Reserve Board had his credit card application at a large toy store chain denied by an automated credit scoring system. The incident opened debate regarding the accuracy, and indeed, the propriety, of scoring systems in the financial services industry. Community development practitioners are concerned about how the industry uses automated credit scoring systems, the risks implicit in these systems, and where the future of these systems may lie. This article addresses some of the challenges posed by credit scoring.

Credit scoring is an objective method for predicting certain credit behaviors of individual borrowers and large populations. Based upon empirical research into the behavior of large numbers of borrowers, credit scoring systems are formulas for predicting such credit behaviors as

  • Who will be late for a payment on a loan? This is usually calculated in the industry parlance of 60 day and 90 day delinquencies.
  • Who will go bankrupt? Financial institutions definitely want to know which customers may cause losses on loans.
  • Who will pay off a credit card account by switching it to another bank? In most cases financial institutions do not want to lose customers to another institution because the cost of replacing the customer is very high.
Credit Scoring

At the heart of the credit scoring system is an odds chart that estimates the percentage of borrowers in a defined population who will default on a loan, cause a loss to the financial institution, or go bankrupt. Scores are generated from analysis of credit bureau data showing borrower payment histories. In many scoring systems, the higher the score, indicating few payment problems, the rarer the odds of a credit problem. The financial institution will use these odds to set cutoff scores delineating the applications they will approve and decline. Applications are run through a formula to arrive at a score for each credit application. Those scores above the cutoff may be approved; those below, may be declined.

Many institutions will have policies for overriding the cutoff, where information not considered in the scoring system is introduced into the ultimate credit decision. Such information may include the applicant's income or the quality of collateral. Highside overrides, causing the denial of an application, may be due to inadequate disposable income. The institution may use lowside overrides to approve applications exhibiting credit problems where collateral is superior.

Uses of Credit Scoring

Credit scoring isn't simply a method for making credit decisions. Financial institutions use credit scoring systems for a wide range of activities.

Loan Collections: Scoring systems are routinely used to manage the loan collection process. Through the analysis of payment histories of large numbers of borrowers, lenders are able to establish effective collection strategies for individual borrowers who have gone past due on their debts. The strategy for each borrower is devised by using scores obtained from credit bureaus as well as scores that embody the borrower's own payment history with the institution.

For example, a typical strategy for a borrower who has an otherwise pristine payment history but who has let his credit card payment go one payment past due, may involve sending just a friendly reminder on his next credit card statement. On the other hand, a similarly tardy borrower with a history of late payments may receive a terse phone call from a collector. Automated scoring systems were developed for the credit card business because the large number of accounts makes the administration of a collections department unwieldy and prohibitively expensive. Furthermore, scoring systems have allowed for the development of collections strategies that are appropriate to the borrower.
Managing Credit Scoring Systems
  • The population of borrowers needs to be large enough to draw valid conclusions
  • Thorough analysis of borrower data is required to construct scoring formulas
  • The scorecards must be applied to appropriate populations
  • Software systems that run the calculations must be periodically checked for accuracy
  • Applicant data must be correctly recorded
  • Results must be monitored for consistency with predictions
  • Scorecards should be reworked as they become less predictive

Credit line adjustments: Periodically, credit card companies will review the credit bureau scores and payment history scores on each account, and raise or lower the credit line for the borrower. An account holder with credit problems may expect to have his credit limit lowered, while one with an excellent payment history and who uses his credit line will likely have the credit line increased. The automated system is necessary to manage the large number of accounts.

Fraud detection: Some of us have received phone calls from our credit card companies asking us if we had actually made certain charges. The credit card companies, reviewing our history of charges, have determined that recent charges do not fit our normal charging pattern. Possibly the charges have been made in a city where we have never made charges. Or maybe some large ticket items such as jewelry and furs have been charged! In order to combat growing credit card fraud, credit card companies have developed sophisticated analytical systems for determining unusual charge activity.

Loan pricing: Traditionally, lenders have set interest rates on consumer loans without regard for the varying risks of the borrowers. Credit scoring of borrowers allows the institution to gauge the risk of default for each borrower, based upon the loss experience of other borrowers with similar credit histories. The institution then translates that risk assessment into an appropriate interest rate. Less risky borrowers pay less than higher risk borrowers.

Marketing: Financial institutions are increasingly using credit bureau scores to select potential borrowers for credit cards, home equity loans, and even mortgage loans. Not only do these scores provide assessments of credit risk, but they also provide insight into potential usage of the credits.

Risks with Automated Underwriting Systems

Automated scoring systems by their very nature rely upon fairly sophisticated algorithms that draw upon large databases. As with all sophisticated and complex systems, careful management is required to obtain the desired results. Below are some critical principles in managing credit scoring systems.
  • The population of borrowers to be analyzed, both good credits and delinquent ones, needs to be large enough to draw valid conclusions. Many financial institutions control this risk by using generic scorecards developed by vendors that draw on large databases.
  • Thorough analysis of the borrower data is required in order to construct the formulas used in the credit scoring systems. Superficial analysis may yield an incomplete profile of the targeted borrowers. Sometimes consultants who have a track record are used to help contain the risks.
  • The financial institution should apply the scorecards to appropriate populations. In most cases the scorecards should only be applied to applicants whose credit attributes are similar to those of the development population of borrowers.
  • Financial institutions must ensure that applicant data is correctly recorded and that the software systems that run the scorecard calculations are periodically checked for accuracy. This is imperative, since software systems are frequently enhanced. A financial institution controls this risk by training operational personnel and auditors to verify the accuracy of credit scoring systems.
  • Financial institutions should routinely monitor the results of the scoring system to ensure that the results are consistent with those predicted by the analysis of the development population. There are a number of standard reports used in the industry for this purpose.
  • As the scorecard becomes less predictive it should be reworked or replaced. Managers of credit risk and internal loan review must maintain a close eye on the scorecard.

All of these risks can be contained with proper management. The better operations have a clear strategy for the use of credit scoring developed through a team approach that involves marketing, credit risk management, and operations staff. The institution's management must assert control over the scoring systems to ensure that risks are fully understood and controlled to the degree necessary to achieve the desired returns. Audit and internal loan review must gain special knowledge of these systems to effectively judge the risk controls. Institutions that use automated scoring systems in a more than superficial manner need to ensure that executive management has a thorough understanding of the risks inherent in these systems.

Why the Increasing Use of Credit Scoring?
  • Recent technological advances allow banks to use huge databases inexpensively.
  • Bank mergers have led to large consumer loan portfolios that require sophisticated management systems.
  • Consumer credit bureau data have become more accurate and more comprehensive than before.
What Does the Future Hold for Credit Scoring?

The use of credit scoring will likely spread into other loan portfolios that have large numbers of accounts, including residential mortgage loans and small business loans. In the past year, the two principal residential mortgage secondary market institutions, Fannie Mae and Freddie Mac, have asked mortgage companies to begin using credit scores in their underwriting of loans.

It is conceivable in the near future that the least risky mortgage loan applications will be approved quickly and with less expense to the borrower. In fact, a mortgage lender in the Atlanta market claims to have made a mortgage loan decision, using an automated scoring system, in less than two minutes! A number of banks are now experimenting with credit scoring of small business loan applications, although there is a consensus that the use of these systems will be limited until the commercial credit bureaus reach as comprehensive a level as the consumer credit bureaus.

Some lenders are experimenting with automated loan machines that allow consumers to obtain unsecured loans in minutes by analyzing credit bureau scores. And, very soon, we will likely see financial institutions, in conjunction with automobile dealers, making nearly instantaneous credit decisions on car loans as automobile dealers use direct access to the financial institutions via the Internet.

Traditionally, financial institutions have used credit bureau data and other consumer information to mass-market products and services. Interestingly, companies are beginning to use the same data to market on an individual basis. This activity is based upon analysis of the spending habits of individuals as reflected by their credit card purchases. In this way, companies can focus their marketing programs directly on consumers who are most likely to use their products and services.


The future of credit scoring, which in today's world of rapidly changing technology looks remarkably like the present, offers society some challenges. As credit scoring is used to underwrite a greater variety of loan products, regulators will have to ensure that federal anti-credit discrimination laws are not violated.

Nearly instantaneous underwriting that forgoes any human review will necessitate greater efforts to guard against the use of inaccurate or incomplete data that may cause the financial institution to make incorrect credit decisions. Of particular concern here is the exclusion of human review that sometimes adds insight into the credit decision that is not provided by a scoring system.

Finally, the growing use of credit scoring raises the issue of personal privacy. Should companies have access to databases containing an individual's credit history and purchasing patterns? And how should that information be used? All of these challenges deserve public discussion, although it is clear that credit scoring systems, when well managed, have the potential to offer financial institutions superior methods for delivering credit products and services.

Robert Kennedy is a senior examiner at the Federal Reserve Bank of Atlanta who examines large banks and bank holding companies. He is currently participating on a Federal Reserve System task force to review issues related to the supervision of retail and small business lending in banks, including credit scoring systems.

Return to Index