Printer Friendly

Fair lending and credit scoring.

There are many legal questions to consider before adopting a credit-scoring system. Lenders looking for a fair-lending safe harbor need to know the issue is nowhere near that simple.

For at least 25 years, ever since credit scoring became widely adopted in the consumer lending industry, mortgage lenders have been speculating about using credit-scoring systems when underwriting mortgage applications. Until recently, neither the means nor the requisite sustained interest existed to convert this idea into reality. Now, however, as the mortgage industry becomes more concentrated in the hands of large, well-capitalized players, the processing of loan applications is becoming increasingly streamlined and computerized. In addition, more and more mortgage lenders are engaging in high-volume operations where speed and the need for efficient data processing are greater than ever before.

Seeing these changes in the mortgage marketplace, the creators of credit-scoring systems are heavily marketing to mortgage lenders. Even credit bureaus are now offering "credit scores," based on the credit histories of individuals in their files. Large investors and mortgage bankers, such as Freddie Mac, Fannie Mae and GE Capital Mortgage Corporation, are seriously exploring the feasibility of using scoring-system technology in purchasing mortgages.

In this environment, the move toward credit scoring seems almost inevitable, despite a residue of cultural resistance on the part of some industry veterans that still see the loan-approval process as more art than science - particularly where real estate collateral is involved. In my view, it's a resistance that can't survive. In reality, single-family residential lending increasingly is being seen as a variant of consumer lending. There seems to be no reason for mortgage lenders not to gain the benefits of the technology that for years has allowed credit card and auto lenders to process millions of applications each year.

However, some caution is in order. As has happened before, oftentimes when private industry moves to innovate, government forces move in inexorably to regulate. Accordingly, mortgage lenders considering adopting some form of credit-scoring technology also should consider the ways in which credit scoring can interplay with fair housing and equal credit concerns, requiring more, not less, attention.

While there has been very little development of case law on the way that credit-scoring systems stack up under the fair lending laws, two things are certain: (1) The move toward credit scoring by mortgage lenders comes at a time when the Department of Justice (DOJ) and Federal Trade Commission (FTC) are stepping up fair lending enforcement to levels not seen in 20 years, and they are beginning to focus specifically on credit scoring in the consumer lending arena; and (2) despite what some credit-scoring advocates may say, simply adopting a system labeled "credit scoring" does not automatically ensure compliance with fair lending laws.

If designed and implemented carefully, credit-scoring systems can have a positive impact on equal credit compliance; but if adopted without critical analysis and individual tailoring, they can seriously complicate a mortgage lenders compliance profile. Credit scoring can come close to being a "safe harbor," but only if you avoid the hidden shoals.

Mortgage lenders are particularly vulnerable

When it comes to government litigation, mortgage lenders have an enormous vulnerability that other consumer lenders do not: an abundance of racially coded data and a history of government enforcement. More than any other factor, the availability of racial data makes mortgage lenders using credit-scoring systems much more vulnerable to an attack based on ostensible "statistical showings" of discriminatory lending results.

Mortgage loan applications, unlike other types of consumer credit, are required to note the race of the applicant. In addition, depository institutions and their subsidiaries are required to keep application data according to census tracts under the Home Mortgage Disclosure Act (HMDA). This permits easy "geocode conversion," using census data that translates readily into neighborhood racial and demographic profiles.

In the absence of intentional, overt discrimination (which nowadays is rare), fair housing lawsuits against lenders almost always depend on an attempt to show a statistical correlation between racial factors and outcomes. In this respect, mortgage lenders that credit score their applications may be early targets of litigation, as the government, like the industry, is beginning to explore the implications of credit scoring.

In addition, unlike many consumer lenders, most mortgage lending institutions (or their affiliates), whether they originate, purchase or securitize loans, are subject to the Community Reinvestment Act (CRA). This means their minority and community-based lending programs are under constant scrutiny by federal regulators and community groups that have learned to use the CRA as a tool in negotiating with expanding financial institutions that sometimes are held hostage by CRA protests.

The Department of Justice has brought several high profile fair-housing and equal-credit cases against mortgage lenders in recent times. DOJ can be expected to continue seeing housing finance as a priority enforcement target. (See, for instance, consent decrees in United States v. Blackpipe State Bank [D.S.D. 1993] [No. 93-5115]; United States v. Chevy Chase Federal Say. Bank & B.E Saul Mortgage Co. [D.D.C. 1994][No. 94-CZ-1829]; United States v. Decatur Federal Sav. & Loan Ass'n. [N.D.G.a. 1992] [No. 92-CV-2198-CAM]; United States v. First Nat'l Bank of Vicksburg, [WD Miss. 1994] [No. 5:94 CV 6(b)(N)]; United States v. Shawmut Mortgage Co. [D. Conn. 1993][No. 3.93 CV-2453AVL]; United States v. American Family Mutual Ins. Co.; [E.D. Wisc. 1995][No. 90-C-0759] United States v. Northern Trust Bank/Lake Forest, NA. et al.; [No. 95.C.3239] [N.D. Ill. June 1, 1995].)

Recently, Justice and the FTC have shown a new eagerness to explore the way the so-called effects test can be applied to credit-scoring systems. The effects test is a judicial rule of interpretation under which a lender, employer, real estate broker or landlord can be held liable even for using facially neutral criteria to qualify applicants where it can be shown statistically that the criteria has a disproportionate impact by race or sex. In fact, in a recent speech before the Independent Bankers Association of America (IBAA), Deval Patrick, assistant attorney general in charge of the Civil Rights Division, made clear his agency's intent to seek out opportunities to apply the effects test aggressively. (See "Patrick Speaks to Business Necessity, Warns on Disclosure of Self-Testing Info," 1995 Daily Report for Executives (BNA), February 16, 1995.)

Even more recently, there have been widespread public reports of a major joint investigation by the Department of Justice and FTC into the consumer lending practices of the three major domestic car company credit subsidiaries. The investigation is believed to focus on credit scoring. (See "Minority Group Wants Justice to Look at All Lenders," Automotive News, March 27, 1995; "Justice Department Probes Fair Lending by Finance Units of Detroit Automakers," Banking Report (BNA), March 27, 1995; and "Lending Probe of Auto Industry Opens New Front in Bias Front," The American Banker, March 23, 1995.)

As to the effects test, Justice has reaffirmed its intent to focus future enforcement activity on "facially neutral practices which result in a metropolitan-wide disparate impact based on race." (See nine-page letter from Assistant Attorney General Deval Patrick to the American Bankers Association and the Mortgage Bankers Association regarding the Department of Justice Fair Lending Enforcement Program, February 21, 1995.)

Finally, when it comes to using credit statistics to "prove" discrimination, the government, as well as private plaintiffs and the media, almost always rely on simplistic gross statistics, rather than sophisticated econometric analysis. Rarely is an effort made to dig more deeply than superficial accept/reject rates to probe the actual basis for credit decisions or to take into account the degree of complexity necessary to avoid statistical showings that are misleading. To be sure, government experts produce reams of computer-generated data. But while this data may have "press release value" and be the basis of forcing creditors into consent decrees, these statistical findings often lack the integrity necessary to hold up in court.

Advantages of credit scoring

With all this in mind, let's look at the advantages and disadvantages of making credit scoring part of the mortgage evaluation process.

The business advantages of credit scoring are plain, and its appeal as a potentially nondiscriminatory credit evaluation device should be obvious.

Purists and advocates of scoring point to potential savings in transaction costs. But more importantly, a carefully constructed credit-scoring system can significantly increase the accuracy of a lender's credit evaluation decision. In other words, credit-scoring systems theoretically give the lender the ability to accept a larger percentage of "good" applications (those that will not go into default or delinquency) and more accurately identify potential "bad" applications. In addition, by raising or lowering a cutoff score, a lender can modulate the amount of credit extended as market circumstances change. Lenders also can extend more credit with a more predictable risk assessment than can be accomplished "by hand."

The implications of a statistical device of this type for securitization and secondary market purposes is also great. The ability to predict performance with greater accuracy translates directly into a more reliable and more marketable mortgage security.

With respect to discrimination, builders of credit-scoring models have long advocated scoring systems over subjective human underwriting evaluation because the system, theoretically, is "blind" to overt or subtle preconceptions about racial or lifestyle factors. In the most simplified form, a credit-scoring system merely "counts" the numbers of "goods" and "bads" in a given population, based on the lender's criteria for performance. The regression analysis correlates and weights performance outcomes with hundreds of objective, demonstrated, historical, creditworthiness factors that have been associated numerically with performance in the past. Where the race of the applicant is actually unknown to the personnel in a remote data processing center, the use of a credit-scoring system allows a lender to defend against allegations of discrimination by showing an absence of any opportunity for discriminatory considerations to come into play. The reliance is on a totally objective and statistically valid set of criteria manifestly related to creditworthiness.

The impact of effects-test principles on credit scoring

Arguments in favor of credit scoring as a tool to preclude the possibility of lending discrimination are challenged by the government's reliance on the so-called effects test as it has developed in the employment discrimination area. The legislative history of the Equal Credit Opportunity Act (ECOA), as well as Regulation B and interpretative materials issued by the Federal Reserve Board, makes clear that the effects test, as it has developed in employment law, is intended to apply to the ECOA. (There are some legitimate questions as to whether it applies in the same way under the Fair Housing Act, but that is beyond the scope of this article.)

We will put aside for now some of the extremely technical issues that might make it difficult for the enforcement agencies, such as Justice and FTC, to apply the effects test in the credit area. We will begin, instead, with a working definition of the effects test. The effects-test rule generally stands for the proposition that even a factor that is "facially neutral" can be prohibited if used in a credit-granting system and it can be shown statistically to have a "disparate impact" on a prohibited basis. If it does, the lender will then be called upon to show that the practice is required as a "business necessity" and (according to the government) that there is no "less discriminatory alternative." Thus, credit-scoring advocates point to credit-scoring systems as helpful to defend against discrimination charges, not only because of the "facially neutral" nature of the factors most often used in credit-scoring systems, but because of the extent to which credit-scoring systems inherently lend themselves to a "business necessity" defense.

For instance, should it be shown statistically that certain factors have a "disparate impact" on account of race or sex because of societal trends beyond the lender's control (i.e., income, net worth, home ownership and so on), the use of a properly built and carefully weighted credit-scoring system should qualify as proof positive of "business necessity" - a system that correlates, weights and balances bona fide and demonstrable creditworthiness factors associated with actual performance over time. (The government may contend that the business necessity defense imposes a higher standard of proof on lenders, but most observers outside the government argue that this position will not hold up if tested in court.)

Many lawyers and compliance experts expect the next wave of equal credit or fair-housing litigation brought by government agencies to be centered in the mortgage lending area where a credit-scoring system is in use and the racial data exists to correlate outcomes to race or neighborhood factors. For instance, many credit-scoring systems rely upon certain occupation codes or definitions of "derogatory" credit or other factors, which may have a high statistical correlation with race or gender. In such circumstances, the government might try to use the credit-scoring system to indict, instead of exculpate, the lender. Should such an attempt be made, the lender will have several avenues of defense.

Problems with applying the effects test to lending

The effects test was developed for cases involving employment discrimination. There are substantial differences, however, between criteria used to prospectively qualify an individual to perform a job and the tens of thousands of pieces of objective historical information about past credit performance processed in a regression analysis. Such analysis is based on past performance that correlate with dozens of objective creditworthiness factors looked at for each individual applicant in a credit-scoring system.

Credit-scoring systems are not "predictive" - they are "associative." The data itself leads to the outcomes, not a subjective, human judgment call about what should or should not "predict" or "cause" creditworthiness. Thus, any effort by the government to apply a superficial effects test analysis to the complex credit-scoring matrix is unlikely to be successful when tested in court under the exacting rules of evidence and cross-examination.

If an effects-test challenge were made to a mortgage lender's use of credit scores, it most likely would be directed at a higher "reject" rate with respect to nonwhites or with respect to collateral located in areas that are predominantly nonwhite. If such a challenge were brought, the government probably would try to point to a prima facie statistical showing of a differential and then try to twist the data into evidence or proof that the credit-scoring system itself is discriminatory.

The argument goes as follows: Because credit-scoring systems are built on populations approved for credit in the past, they are programmed inherently to look only at the financial and demographic characteristics of those populations, which may not be sufficiently inclusive. To the extent that a credit-scoring system is built on a population of past borrowers who are predominantly white or suburban, the system arguably will have a built-in bias, because it will fail to fully identify those factors that correlate with creditworthiness in different, more urban or ethnic populations.

It is fair to say that this argument pushes the envelope - it represents the furthest possible extension of a tautology that the courts are unlikely to accept. (See, for instance, Cherry v. Amoco Oil Co., 490 F. Supp. 1026, 1031 n.9 [N.D. Ga. 1980].) Under this reasoning, the very statistics that demonstrate broad societal correlations between race or gender for such things as income, homeownership, number of credit cards or even credit performance would be off-limits to lenders, when, in reality, they are the best, most reliable and most neutral predictive factors for future credit performance. For instance, statistics may show that a particular group has a lower incidence of bank accounts or credit cards than the population as a whole, and therefore, scores more poorly. Does this make a lender's reliance on credit cards and bank accounts discriminatory?

Unlike the employment area, where height, weight or possession of a high school diploma may have little causative or statistical correlation with future performance in a particular job category, the "pertinent elements of creditworthiness" about an individual applicant used by most credit-scoring systems can be shown to comprise the very essence of factors that predict credit performance.

Things to consider

There are a number of things that mortgage lenders contemplating the use of credit scoring should do to maximize their ability to defend their systems.

When evaluating the adoption of a credit-scoring system lenders should consider the following:

* Use credit scoring as a guide. Credit-scoring purists maintain that there should be no manual overrides or subjective human evaluation added to a system that relies on credit scoring. On the other hand, the strict application of numerical cutoffs could result in at least anecdotal instances of perceived discrimination because of a lack of flexibility. Accordingly, where an otherwise well-qualified applicant has an explainable credit history problem caused perhaps by illness or temporary layoff, pure adherence to credit-scoring criteria might reject that individual. But a procedure for manual review might identify instances where flexibility is preferred.

At the same time it can be argued that by introducing flexibility a lender invites its employees to exercise flexibility on a subjective, and possibly discriminatory, basis (i.e., by bending the rules for members of some groups but not for others).

Many lenders, even in nonmortgage consumer lending areas, are most comfortable with a system in which the credit score is used as a guide - albeit an important one - particularly at the high and low ends of qualification. In these systems, lenders use a degree of flexibility and manual review for those who score in the middle range. Where exceptions are to be made, they should be made only pursuant to a (preferably) written, objective set of criteria that define the various bases for exceptions and flexibility.

* The hazards of a borrowed system. Regulation B, issued by the Federal Reserve Board under ECOA, permits lenders to borrow a credit-scoring system built on someone else's population in certain circumstances. However, this does not insulate the borrowing lender from a credit-scoring system that is poorly constructed; a system that inadvertently is built on a skewed population unrepresentative of the lender's own application field; or the need to validate a system based on the lender's own experience over time.

In addition, many lenders that use borrowed systems, including those provided by the credit bureaus, do not realize that it is the lender's obligation to provide an accurate reason for adverse action if that reason relates to a low credit score. The rules for disclosing adverse action on the basis of credit score are complex, and most lenders that use scoring systems have a computerized system in place to provide this disclosure. Lenders using a borrowed system may not realize that although the credit bureau is not obligated to disclose a credit score when an inquiry is made under the Fair Credit Reporting Act, lenders do not get the benefit of this exemption. Thus, if you're going to use a credit score purchased from a credit bureau, don't do so without preparing to disclose credit score information in compliance with Regulation B to your own turned-down applicants.

* Look at the score card. The architecture, design and implementation of a credit-scoring system is a science. In fact, a credit-scoring system must meet certain carefully defined minimum parameters to be considered a statistically valid and empirically derived credit system under Regulation B.

If the system is not qualified on a technical basis, the lender loses certain critically important safe harbor protections, particularly when it comes to the use of age in extending credit. The mere assignment of numbers to credit variables or use of the label "credit score" does not constitute a credit-scoring system that will hold up under Regulation B or the scrutiny of judicial analysis in the event of an effects-test challenge.

Accordingly, no mortgage lender should adopt a credit-scoring system without carefully investigating how it was built. It is the lender that will bear the consequences of a poorly built and executed system. There are several key questions to ask. What does the score card look like? Are you comfortable defending each characteristic and attribute that is scored? Was the system built on stale or current data? Whose data was used? What type of credit was involved? What was the source and breadth of the application flow involved? Has the credit-scoring builder given you appropriate warranties of compliance with the ECOA, including indemnification of legal defense costs? Does the vendor have the required expertise and experience in performing the complex computer function necessary to build a system?

* Do not assume that use of a credit-scoring system is an automatic safe harbor. If you are a mortgage lender that has a racially and economically diverse application flow, what impact will there be if you use a credit-scoring system designed principally on a population borrowed from a high-income travel and entertainment card credit operation? Are the factors that predicted delinquency among the borrowed population the same as those associated with mortgage delinquency? (How do you add in the fact that people tend to pay their mortgage first, which may not be reflected in the performance of the borrowed population?)

Do you need to validate a system based on your type of credit or population? Are your advertising, marketing, application and prescreening procedures creating a discriminatory application pool before the credit-scoring system even comes into play? Have you tested the credit-scoring system against your own population to see how it fared in predicting both "goods" and "bads?"

These are all important self-diagnostic questions to ask. They also are questions you should have the answers to - before the government comes knocking on your door.

Warren L. Dennis is a partner at the law firm of Proskauer Rose Goetz & Mendelsohn LLP in Washington, D.C.
COPYRIGHT 1995 Mortgage Bankers Association of America
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1995 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:mortgage lenders
Author:Dennis, Warren L.
Publication:Mortgage Banking
Article Type:Cover Story
Date:Nov 1, 1995
Words:3612
Previous Article:The debate over credit scoring.
Next Article:Dealing with an epidemic.
Topics:


Related Articles
Scoring borrower risk.
No panacea.
The debate over credit scoring.
Credit risk, credit scoring, and the performance of home mortgages.
A cloud over the business.
Shedding new light on credit scoring.
The HMDA controversy: the Federal Reserve this month releases a summary of data from lenders showing pricing differentials sorted by ethnic and...
HMDA revisited: the date is fast approaching for release of Home Mortgage Disclosure Act data for 2005. Make sure you are ready to explain what the...
Paul F. Hancock--mortgage industry attorney and fair-lending expert.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters