Staff Training, Loan Pricing and Data Accuracy

Credit scoring is an underwriting tool used to evaluate the creditworthiness of prospective borrowers. Used for several decades to underwrite certain forms of consumer credit, scoring has become common in the mortgage lending industry only in the past 10 years. Scoring brings a high level of efficiency to the underwriting process, but it also has raised concerns about fair lending among historically underserved populations.

The mission of the Federal Reserve System's Credit Scoring Committee is to publish a variety of perspectives on credit scoring in the mortgage underwriting process, specifically with respect to potential disparities between white and minority homebuyers. To this end, the committee is producing a five-installment series of articles. The introductory article provided the context for the issues addressed by the series. The second article dealt with lending policy development, credit-scoring model selection and model maintenance. The third article explored how lenders monitor the practices of their third-party brokers, especially for compliance with fair-lending laws, pricing policies and the use of credit-scoring models.

The fourth article focuses on staff training, the level and consistency of assistance provided to prospective borrowers and the degree to which applicants are informed about the ramifications of credit scoring and data accuracy in the mortgage application and underwriting process.

Representatives of three organizations were asked to comment. They were selected because of their different perspectives on credit scoring and fair lending.

William N. Lund
Maine Office of Consumer Credit Regulation
Mr. Lund is director of Maine's Office of Consumer Credit Regulation. A graduate of Bowdoin College and the University of Maine School of Law, he worked in private practice and with the Maine Attorney General's Office prior to assuming his current position in 1987. Mr. Lund has served as chair of the Federal Reserve Board's Consumer Advisory Council. He writes and speaks frequently on consumer law issues.

John M. Robinson III and Ken Dunlap
Midwest BankCentre
Mr. Robinson is the audit director/compliance officer and Community Reinvestment Act officer for Midwest BankCentre in St. Louis. Robinson has 16 years of banking experience with the last 10 in internal audit and compliance management. He is a graduate of Westminster College, of Cambridge University's master's program and of the American Bankers Association's National Compliance School. He is chairman of the Missouri Bankers Association Compliance Committee and a board member and speaker on compliance topics for the Gateway Region Center for Financial Training.

Mr. Dunlap is the loan processing manager/chief underwriter for Midwest BankCentre. Dunlap has 11 years of experience in mortgage underwriting, compliance, Home Mortgage Disclosure Act and Community Reinvestment Act reporting, and loan platform maintenance. He is a graduate of Southeast Missouri State University and has been with Midwest BankCentre for five years.

Midwest BankCentre, a $720 million community bank with nine offices, was named "Outstanding Small Lender" by the Small Business Association in 2000. The bank originated 569 mortgage loans in 2000 and had a higher percentage of mortgage home improvement loans in compliance-sensitive segments than did its community peer banks. Midwest BankCentre was instrumental in helping to establish the Lemay Housing Partnership.

Josh Silver
National Community Reinvestment Coalition
Josh Silver has been the vice president of research and policy at the National Community Reinvestment Coalition (NCRC) since 1995. He has a major role in developing NCRC's policy positions on the Community Reinvestment Act (CRA) and other fair-lending laws and regulations. He has also written congressional testimony and conducted numerous research studies on lending trends to minority and working class communities. These studies include NCRC's Best and Worst Lenders, a comprehensive analysis of home lending in 20 metropolitan areas, and a report sponsored by HUD on the performance of Fannie Mae and Freddie Mac in financing home loans for minority and low- and moderate-income borrowers. Prior to joining NCRC, Mr. Silver was a research analyst with the Urban Institute. Mr. Silver holds a masters' degree in public affairs from the Lyndon B. Johnson School of Public Affairs at the University of Texas in Austin and a bachelor's degree in economics from Columbia University. NCRC is the nation's CRA trade association of more than 800 community groups and local public agency member organizations. For more information about NCRC, call (202) 628-8866 or visit the coalition's web page at http://www.ncrc.org.

The contributors to this article were asked to respond to the following statement:

In the past, the terms "thick file syndrome" and "thin file syndrome" were used to describe the allegation that white and minority mortgage applicants received differing levels or quality of assistance in preparing mortgage applications. These terms were used primarily before the advent of credit scoring in mortgage lending. In the current mortgage market environment, credit and mortgage scoring have taken a front seat to judgmental systems. With greater reliance on these automated systems and less human judgment in the decision process, the quality of assistance provided applicants is even more important.

Given the increased reliance on automated underwriting, what should lenders do to ensure that:

• Lending policy is strictly observed and that any assistance offered to loan applicants or prospective applicants to improve their credit score is offered equitably.
• Applicants have a clear understanding of the importance of their credit score to the approval and pricing processes.
• Staff training and oversight regarding credit policy and fair lending guidelines are adequate to ensure consistent and fair treatment of loan applicants.

Response of William N. Lund
Maine Office of Consumer Credit Regulation

As a regulator enforcing Maine's credit reporting laws, I have tried to learn as much as I can about credit scoring. The ingenuity of the scoring models and the complexity of the applied mathematics are very impressive, and I have no doubt that use of such scores permits creditors to make fast decisions on consumers' applications. However, from the consumer's perspective, I harbor great concerns about the exponential growth in the use of such scores. I can summarize my concerns as follows:

Concern #1: Credit scoring has led to a "re-mystification" of the credit reporting system.
In 1969, during the debate on the original Fair Credit Reporting Act (FCRA), Wisconsin Senator William Proxmire spoke of the congressional intent behind the law:

"The aim of the Fair Credit Reporting Act is to see that the credit reporting system serves the consumer as well as the industry. The consumer has a right to information which is accurate; he has a right to correct inaccurate or misleading information, [and] he has a right to know when inaccurate information is entered into his file…. The Fair Credit Reporting Act seeks to secure these rights."

In other words, passage of the FCRA represented an effort to "de-mystify" the credit decision-making process. In the years since passage of the act, consumers and creditors have become relatively comfortable with the use of traditional credit reports.

However, I fear that the creation and use of credit scoring systems constitutes a step backward from the goals of the Fair Credit Reporting Act to make credit reporting data accessible, understandable and correctable, and to make credit reporting agencies responsive to consumers. In other words, just as the FCRA "de-mystified" the storage and use of credit information, credit scoring is now serving to "re-mystify" that process.

Concern #2: A double impact results when an error in the underlying data impacts a credit score.
The fact that a large percentage of credit report data is accurate is of little comfort to a consumer whose report contains harmful errors. If errors in the underlying data result in a low credit score, in effect the original error is compounded.

In addition, the consumer now finds himself twice removed from the actual problems. A credit-scoring system creates a new layer of data, and that new layer separates the consumer from the raw data. The system as a whole becomes less accountable to consumers. When the Federal Trade Commission decided not to treat credit scores the same as traditional reports, not only did this decision remove the legal responsibility to disclose the score but also to correct an inaccurate score and notify previous recipients at the consumer's request.

Concern #3: Because there are so many different products, and because these products are ever-changing, consumers cannot be educated about common rules or standards.
Let's look at the current range of products: Trans Union has Emperica, Experian uses the name Experian/Fair Isaac, and Equifax offers Beacon. In addition, Fannie Mae has developed Desktop Underwriter (DU), while Freddie Mac uses its Loan Prospector. Other lenders use Axion or Pinnacle.

Over the years, those of us who assist consumers with credit report issues have managed to get our arms around the "big three," but it is much more difficult to make sense of the myriad variations on the credit-scoring theme. Even something as simple as score values is very confusing: My files contain the statements of four different experts who describe the range of scores in the basic Fair, Isaac (FICO) model as 300 to 900, 400 to 900, 336 to 843, and 395 to 848. If products offerings are such that the "experts" can't agree on basic information, how can consumers be expected to gain a meaningful understanding of the scoring process and its impact?

Concern #4. Reason codes: Everyone gets four, regardless of how good or bad their scores.
For those with great scores, four may be too many. For those with low scores, four may be too few.

Why can't reason codes be specific, as in, "The fact that your 1972 Pinto was repossessed in January results in a reduction of about 40 points from your score." Don't we have the technology to do that?

In addition, some of the factors used to determine scores seem illogical on their face, the most obvious being the effect of closing existing, older, unused credit accounts. From most real-life perspectives, closing such accounts should be a good thing. From a scoring perspective, however, that action harms a score in two ways: First, it increases the ratio of used credit to available credit, by reducing the denominator of that fraction. Second, it decreases the average age of a consumer's credit lines, resulting in further score reduction.

As another example, industry sources have told me that a consumer gains points for doing business with established banks, but loses points for doing business with small loan companies or check-cashers, even if payment histories are identical. In other words, there is good credit and bad credit, which may have more to do with a consumer's neighborhood and lifestyle than with an accurate prediction of the chances of future repayment.

And consider the advice that consumer advocates have given for years: Compare APRs and shop around for credit to get the best deal. Shopping around these days means piling up inquiries on one's credit report. Despite recent efforts within Fair, Isaac-based models to discount groups of inquiries, the fact remains that inquiries form a component of a credit score.

The use of credit scores for non-credit decisions adds to the illogic. For example, should paying cash for purchases result in an increase in a consumer's auto insurance rates? That is the outcome when a "thin" file results in a low credit score which is used by an insurer to set premiums.

Concern #5: Creditors will probably begin to rely too heavily and exclusively on credit scores, despite "instructions" to the contrary.
What was introduced as a tool to be used in conjunction with other criteria is quickly becoming a litmus test. Creditors are busy, and underwriters are often not rewarded for taking risks. The logical result will be a dependency on credit scores and a reluctance to look to a broader picture. To quote Chris Larsen, CEO of online lender E-Loan: "Lenders are increasingly relying on these scores. Many loan products, including some home equity loans and auto loans, are based almost entirely on your FICO score."

Conclusion
Many aspects of the credit scoring process have now gotten ahead of the ability of consumers to make sense of the system and of regulators to meaningfully assist those consumers. Providers of credit scores should be required to share responsibility for ensuring the accuracy of the underlying data, of correcting that data and of disseminating the correct information if requested by the consumer. Despite repeated assertions by the industry that credit scoring is not a mysterious black box, the lack of any uniformity, oversight or accountability make that analogy too close to the truth.

Response of John M. Robinson III and Ken Dunlap
Midwest BankCentre

Given the increased reliance on automated underwriting, what should lenders do to ensure that their lending policy is strictly observed and that any assistance offered to loan applicants or prospective applicants to improve their credit score is offered equitably?

Lending policies must be observed to ensure sound financial business decisions and to avoid any potential disparate treatment of applicants. At the same time, policies must allow lenders to evaluate individual credit needs and varying applicant scenarios. Lenders must be conscious of nontraditional applicants for whom relaxed underwriting may be key in obtaining a loan. For example, Midwest BankCentre offers the FreddieMac Affordable Gold "97" mortgage product for first time home-buyers. This program, in contrast to many others, allows for a 3% down payment from any source (e.g., gifts).

How a mortgage credit decision is made is one of the two keys of potential discrimination. Prescreening is the other. Underwriting standards and policy adherence are very important. Allowing excessive overrides creates an atmosphere for potential discrimination-when a lender decides to override an established and proven underwriting decision, the reason is personal more times than not. Banks should have workable, clearly written policies and underwriting guidelines. Every lending decision should be fully and clearly documented, especially if a lender overrides a prescribed credit score and makes the loan. Lending institutions must give equal assistance to all applicants.

To avoid problems with loan policy standards, the following steps should be taken:

• Review bank policies and procedures. Compare them with actual file reviews.
• Review all underwriting and credit score overrides. Look for patterns.
• Review loan files and denials for adequate documentation. Look at all forms, documents and disclosures in the files.

Given the increased reliance on automated underwriting, what should lenders do to ensure that their lending policy is strictly observed and that any assistance offered to loan applicants or prospective applicants to improve their credit score is offered equitably?

Generally speaking, the average mortgage applicant - especially the first-time home buyer - does not understand clearly how a credit score affects the mortgage outcome. Applicants who have never had a loan or a problem with a loan decision probably have never heard of a credit score. Knowing how to use a credit score involves knowing what is in the score and what it does and does not tell about the prospective applicant. Because the score is based on data provided by a credit bureau, applicants should be instructed on how to rectify any error or problem that appears on their credit bureau reports.

If a bank or creditor does not use a credit bureau service, then the applicant's credit history is not recorded. These scores do not reflect information such as the amount of down payment, income, cash flow, or other mitigating assets. The score is only part of the applicant's credit picture. Therefore, one may conclude that too much reliance on credit scores or on automated decisions could raise flags of disparate impact issues. In actuality, there may be many reasons why a low score would not be a negative in the bank's decision. For example, a large down payment or significant cash flow could justify overriding a low score. We do make loans to applicants who may not have stellar credit-Freddie Mac guidelines allow for A- offerings-but the interest rates are usually higher.

Given the increased reliance on automated underwriting, what should lenders do to ensure that staff training and oversight regarding the credit policy and fair-lending guidelines are adequate to ensure consistent and fair treatment of loan applicants?

First, all lenders in the bank should know the products offered and always explain to prospective applicants the loan product choices and their associated potential costs. We need to take our responsibility to customers seriously. We earn the trust of customers by how we treat them.

Lenders using their own instincts instead of a score have a different perspective on customer relationships. When looking at the overrides in credit scores, management should look at the decisions made and where and by whom (which branch/lender). Management should look at patterns and at loans that have gone bad and compare them with any initial credit score. Self-testing and self-analysis with an eye on patterns and trends related to any disparity are vital to the organization.

Lenders should follow these basic steps:

• Disclose and explain any conditions for a product or service as well as the benefits of each one.
• Offer the same product to everyone who has comparable qualifications.

To ensure fair and equal treatment of all customers in the application of our credit policies, Midwest BankCentre's compliance department holds annual mandatory fair- lending and diversity awareness training seminars for staff. The sessions are intended to generate discussion about how well employees understand fair-lending laws and issues of cultural diversity in the workplace. We use a video titled True Colors, the ABC Prime Time Live telecast filmed on location in St. Louis, and each attendee receives the booklet Closing the Gap - A Guide to Equal Opportunity Lending, published by the Federal Reserve Bank of Boston. We have also used other videos from corVISION Media Inc.-in particular, Valuing Diversity at the Interpersonal Level. Participants complete and discuss a self-assessment checklist that underscores their own perceptions of understanding differences and adopting changes.

Being a community bank, we do not rely heavily on credit scoring; we still consider the individual borrower's overall credit reputation. Because we continue to have direct interaction with our applicants throughout the credit process, it is important that our mortgage lenders receive ongoing training in what constitutes fair and consistent treatment.

Response of Josh Silver
National Community Reinvestment Coalition

Toward meaningful disclosure and discussion of credit scores

All of us have credit scores, but most of us don't know what they mean. If we knew what they meant, would we be more likely to get approved for a low-cost loan? The answer is probably, but the disclosures of credit scores have to be meaningful if they are to be helpful to the borrower.

Credit scores are numbers ranging from 300 to 800 that are supposed to reflect the risk that we, as borrowers, pose to banks. The higher the score, the less risky we are and the less likely we will be late on loan payments or default on the loan altogether. Credit scores are calculated on the basis of a credit history that is collected and stored in three major credit reporting agencies or private sector credit bureaus. The record of paying on time or paying late, the amount of debt compared with the amount of available credit on credit cards, and the length of time using credit are major factors that contribute to the score.

If a borrower has a score above 660, he most likely will qualify for a prime rate loan at interest rates advertised in newspapers. If a borrower has a score significantly below 660, he is likely to receive a subprime loan at interest rates ranging from 2 to 4 percentage points above widely advertised rates. The rationale behind the higher rate on subprime loans is that the bank is compensated for accepting the higher risk of delinquency and default associated with lending to a consumer with blemished credit.

Credit scores have been used for decades for consumer and credit card lending. In the mid-1990s, credit scores became a widely used tool in mortgage lending as well. It is not the only criterion banks and mortgage companies use, but it is an important criterion, ranking up there with loan-to-value ratios and total debt-to-income ratios. Proponents of credit scoring assert that its use has increased lending to minority and low- and moderate-income borrowers because it is an objective assessment of a borrower's creditworthiness: Subjectivity is removed from the loan process, and the chances of discrimination are decreased. It is further claimed that credit scoring makes the loan process much more efficient and saves resources that can be devoted to carefully analyzing marginal cases.

The National Community Reinvestment Coalition (NCRC) does not believe that credit scoring has revolutionized access to credit, and neither has the advent of subprime lending, for that matter. Instead, the strengthening of the Community Reinvestment Act (CRA) and the stepped up enforcement of fair-lending laws have been the major forces behind the explosion of credit for minority and low- and moderate-income borrowers during the 1990s. Lenders made only 18 percent of their home mortgage loans to low- and moderate-income borrowers in 1990. The low- and moderate-income loan share surged 8 percentage points to 26 percent by 1995, but by 1999 it had climbed only 3 more percentage points, to 29 percent.

Let's review the major events coinciding with the big jump in lending during the first part of the 1990s and the major events during the lending slowdown in the second half. Congress mandated the public dissemination of CRA ratings in 1990 and the improvement of Home Mortgage Disclosure Act (HMDA) data to include the race, income and gender of the borrower. In 1995, after a highly visible and lengthy review process during previous years, federal banking agencies strengthened CRA regulations to emphasize lending performance as opposed to process on CRA examinations. During the same time period, the Justice Department settled several fair-lending lawsuits with major lending institutions. After 1995, the mortgage industry widely adopted credit scoring, and subprime lending took off. Home mortgage lending increased in the first part of the decade as policy-makers strengthened and applied CRA and fair-lending laws. Lending slowed down in the second half of the decade; during this period, credit scoring and subprime lending were on the rise. Economic conditions played less of a role in the different trends in lending because we were blessed with a tremendous economic recovery during the entire 1990s.

The reason credit scoring was not responsible for the explosion of home mortgage lending to low- and moderate-income borrowers is that credit scoring is not designed to serve those who have the least experience with the financial industry. Credit scoring depends on an established credit history, so that econometric equations can judge the odds of a borrower paying late or defaulting. Officials at one large bank NCRC interviewed for this article stated that they do not use credit scores in their approval decisions regarding special affordable loan programs. They indicated that those people among the low- and moderate-income population who are targeted by special affordable loan programs have low credit scores because they do not have much of a credit history. Instead, the bank uses nontraditional credit history, such as evaluating the timeliness of rent and utility payments. It is likely that CRA encouraged this bank to establish the special affordable loan programs. For this large bank, and probably for many other banks, CRA has more to do with increasing lending to low- and moderate-income borrowers than credit scoring.

Why disclosure would help

While credit scoring has not had a noticeable impact on increasing credit to traditionally underserved borrowers, meaningful disclosures of credit scores would nevertheless help increase access to affordable credit. The optimal time for disclosure is before a customer applies for a loan. If a customer obtains a credit score and the major factors for that score before reaching the loan application stage, he would have a good idea of his creditworthiness. The customer would be in a better position to know if he was getting a good deal on the loan or whether to bargain with the lender.

The caveat is that a consumer must have a clear understanding of what the credit score is and what factors affected his score. The disclosure of the number itself has little meaning. If the credit score is low, for example, the consumer needs to know which factors in his credit history had the most impact on lowering the score. He could then decide whether to delay applying for the loan and how best to clean up his credit. For this reason, HomeFree-USA, a counseling agency in Washington, D.C., and a member organization of NCRC, always includes credit score counseling in its homebuyer preparation courses. Similarly, NCRC educates consumers about their credit scores in its financial literacy curriculum.

Although credit scores are imperfect estimators of creditworthiness, disclosure of credit scores can help reduce the incidence of discrimination in prices, particularly in the area of subprime lending. Fannie Mae's chief executive officer has been quoted as saying that 50 percent of subprime borrowers could have qualified for lower rates. Freddie Mac issued a statement on its web page a few years ago saying that up to 30 percent of subprime borrowers could have qualified for lower-priced credit. A paper commissioned by the Research Institute for Housing America concluded that after controlling for credit risk, minorities were more likely to receive subprime loans.

An unanswered question is how many borrowers who were inappropriately placed into the subprime loan category could have avoided this if they had simply known about their credit scores. Also, how many of them could have obtained lower interest rate loans, even if the loans remained subprime? For example, if an educated borrower knew that his score was 620, which is generally considered A- credit, and was quoted an interest rate 4 percentage points higher than the widely advertised rate, he would know that he was being overcharged. While other underwriting factors, such as loan-to-value and debt-to-income ratios, also contribute to the pricing decision, meaningful credit score disclosures alert borrowers when quotes are or at least seem far higher than they should be.

As California was passing a law requiring credit bureaus to disclose credit scores, Fair, Isaac and Co. Inc., one of the major firms producing scores, took a constructive step and made credit scores available for a small fee through its web site, myfico.com. The company also has a description on its web page of the major factors influencing the score and the weight of each factor.

How banks should disclose and use credit scores

The new California law also requires banks to disclose credit scores to consumers applying for loans. California is the only state to require this disclosure. Several bills working their way through Congress would also require credit bureaus and banks to disclose credit scores.

For the consumer, it is advantageous to be armed with credit score information and to take action to improve the score, if needed, before applying to a bank. However, if a consumer does not have a credit score prior to application, disclosure by the lending institution is still valuable. In a loan approval decision, for example, disclosure of the credit score will help the borrower understand why his loan had a certain interest rate. If the interest rate is in the subprime range, the borrower may want to take steps to improve his credit before closing on the loan. In the cases of loan denial, a lender is required under the Equal Credit Opportunity Act to send a borrower an "adverse action notice." If the reason for the rejection involved one of the factors in a credit score, that factor must be discussed in the adverse notice.

Lending institutions can run afoul of fair-lending laws quickly if they are not careful about using credit scores when helping borrowers apply for loans. For example, in 1999, the Department of Justice settled a fair-lending lawsuit with Deposit Guaranty National Bank over Deposit Guaranty's alleged arbitrary and discriminatory use (or disregard) of credit scores. The lawsuit came about after an examination by the Office of the Comptroller of the Currency concluded that Deposit Guaranty disregarded credit scores when approving loans for whites but rejected blacks with similar credit scores. As a result, the black rejection rate was three times the declination rate for whites.

It is important and valuable for a bank to institute a review process for declined applicants, especially those on the margins of approval. Such a review process may help banks make more loans to minority and low- and moderate-income applicants with little traditional credit history. A judgmental review process must establish consistent criteria by which to overrule credit scores. Such criteria can include consideration of nontraditional credit, including rental and utility payment histories.

Disclosure with a twist

The NCRC believes that information in the HMDA data about credit scores could be instrumental in resuming steady increases in access to credit for minority and low- and moderate-income borrowers. Several months ago, the Federal Reserve Board asked for public comment on its proposal to include the annual percentage rate (APR) in HMDA data.

In response to the Federal Reserve's proposal, NCRC pointed out that the APR, along with credit score information, could vastly improve our knowledge of how credit scores impact pricing and approval decisions. Because many kinds of credit scores exist, it would be difficult to interpret what actual numerical scores mean if they were added to HMDA data. At the very least, the loan-by-loan data could indicate if a credit-scoring system was used and the type of credit-scoring system, such as a bureau or custom score. Policymakers would then have important insights as to whether most loans to minority and low- and moderate-income borrowers are credit-scored and whether banks using credit-scoring systems are more or less successful in approving loans to traditionally underserved borrowers. Community groups and counseling agencies could then use this additional information in HMDA data in their advice to borrowers about which banks are most likely to use credit-scoring systems in a fair manner to provide loans at reasonable rates.

Conclusion

In announcing a Bush administration proposal to provide the public with data on the quality of nursing homes and Medicare health plans, Thomas Scully, a senior official at the Department of Health and Human Services, stated: "Collecting data and publishing it changes behavior faster than anything else." The motivational force of data disclosure under CRA and HMDA has helped activists and the public at large work with banks to increase lending to minority and working class borrowers. Meaningful disclosures of credit scores to consumers and incorporating credit score information in HMDA data would be two more valuable tools for building wealth in traditionally underserved communities.

Previous | Top | Next