In the Press

Mortgage Banking

The Thin-File Problem

March 1, 2007

The lack of sufficient credit information to produce a traditional credit score is preventing some borrowers from becoming homeowners. New advancements with alternative scoring technology hold great promise for a whole new population of borrowers. With mortgage volume having slowed dramatically, there is renewed interest today in being able to expand the pool of potential home- loan borrowers. In particular, the credit industry, perhaps feeling pressure from lenders or even regulators (but also seeing opportunity), has become very proactive in expanding credit models and thus allowing increased lender penetration into the minority and new-immigrant population sectors. The problem in the past has been that as many as 35 million Americans who could potentially qualify for credit from mainstream lenders have been excluded because they lacked traditional credit histories, according to the Washington, D.C.-based Political and Economic Research Council (PERC). Michael Turner, president of PERC, says that while consumer credit- reporting has become a principal means of risk assessment for gaining access to the credit markets, consumers who have little or no information in their credit files - although they may be creditworthy-are actually deemed not creditworthy. To address this issue, credit-scoring companies, as well as some private lenders, have come up with a wide range of solutions to broaden access to that swath of the population either grouped as "thin-file" (having little information in their credit file) or those having no files at all.

Crux of the matter

Nearly 18 million Americans have files too thin to produce a credit score, and another 17 million have no files at all, reports Experian Americas, the Costa Mesa, California-based arm of Dublin, Ireland-based Experian Group Limited. In a combined report by PERC and Experian released in July 2006, it was reported that thin-file and no-file groups were concentrated among minorities (notably African Americans and Latinos), those with low income, the elderly, recent widows and new immigrants. In addition, the report stated, of the approximately 10 percent of households in the United States that do not have a bank account, more than half of those are African American or Latino. Of the 210 million adults in the United States, the credit- underserved market consists of anywhere from one-fourth to one- third of that population. The National Credit Reporting Association Inc., Bloomingdale, Illinois, counts 70 million as credit- underserved, while a more optimistic figure of 50 million is used by Fair Isaac Corporation, the Minneapolis firm that develops credit- scoring systems, in particular the widely used FICO score. According to Fair Isaac's data, 15 percent of the country's adult population falls into the thin-file category (that would earn no FICO score), while 10 percent would garner no "hits" whatsoever (again, no FICO score).

Feeling left out

Fair Isaac builds its scoring systems from data supplied by the three national credit repositories: Experian; Equifax Inc., Atlanta; and TransUnion LLC, Chicago. But what happens when there are 50 million consumers over the age of 18 who haven't had any credit yet, and therefore nobody is reporting information about them to the credit bureaus? "And this is the population from which the lenders expect to experience awesome growth opportunities going forward," says Lisa Nelson, Fair Isaac's vice president of product management. "Thin- file and no-hit represent 25 percent of the U.S. adult population, and there is not enough data about them in national credit repositories to generate a FICO score." Who falls into this category?, Nelson asks rhetorically. "Young adults, recent immigrants, older adults who have not historically used credit and even third-generation immigrants who have not traditionally used credit," she says. In a May 2006 report, the Brookings Institution, Washington, D.C., took a look at credit scores on a county-by-county basis because, as it noted, "consumer credit scores are playing a growing role in the economic mobility of consumers today." The report noted, "counties with relatively high proportions of racial and ethnic minorities are more likely to have lower average credit scores." Matt Fellowes, a Brookings Institution fellow and author of the study, concluded, "This evidence does not suggest that a bias exists." The fact that some minorities use less credit and participate in credit markets less than whites or other ethnic groups would to some extent mean thinner credit files, but "the extent to which thin credit files drive down credit scores is still to be determined," explains Fellowes. Efforts are being made by Fair Isaac and the three major credit repositories to address the issue that there is a whole lot of information out there today that is not being collected about consumer behavior and expectations. This suggests that if you had that information, "you would have a better ability to predict someone's behavior in the future," says Fellowes. "The motivation is not to improve people's credit scores, but to improve the bureaus' ability to predict someone's behavior." Low-income populations and some emerging ethnic groups have become huge markets for the mortgage industry. However, mortgage lenders have not been able to serve these groups efficiently because the existing methodology used to assess risk (automated underwriting systems) is not tailored to these newer target consumer groups. One answer to this problem has been to try to incorporate nontraditional data such as consistent rental, telecommunication and/ or utility bill payments as an alternative means to measure creditworthiness.

Taking another look

In 2006, PERC undertook what it considered a three-part (quantitative analysis, qualitative analysis and an active or alpha phase) look at credit scoring. By the end of the year, it was able to report on the first two phases. In the quantitative and qualitative analyses, the report dealt with alternative data that would allow thin-file and no-file consumers to gain affordable, mainstream credit. Among the various types of transactions looked at were auto insurance, electric and utility bills, rental payment data and telecommunications bills. Rental data were eventually ruled out as the information was too fragmented, so the study focused on payment records for utilities of various types (gas, electric, oil and water) and telecommunications. "Basically, we decided the impact on including these alternative data sets would be the greatest on unscorable and thin-file Americans, and the net impact should be acceptable in terms of score distribution impact," reports PERC's Turner. "We ran simulations through 10 models, then we redacted the nontraditional data and ran the same models and compared the difference. We saw that including alternative data significantly reduced the thin-file population. In our sample, the thin-file population was reduced by 31 percent." Of the 9 million credit files used in the study, 17 percent were considered thin-file. But when just using utility alternative data, the thin-file group dropped to 12 percent. In regard to score distribution, originally about 17 percent of the thin-file group scored above 620, or the typical subprime mortgage cutoff level, says Turner, but when alternative data were included, 51 percent of the thin-file group scored above the subprime cutoff level-"which presents a very different picture for lenders," he says. The study also added demographic information to the credit-file data. This resulted in Latinos being the largest net beneficiaries in terms of an increase in credit scores, with 22 percent more able to get access to credit; 21 percent for African Americans; 14 percent for Asians; 14 percent for those aged 25 years or younger; 14 percent for those 66 years of age; and 21 percent for those who earn $25,000 or less. "The original assumption among mainstream lenders is that little information meant high risk," explains Turner. "They have to take that position, because they rely on data tools. So for them to be able to grow into new markets, particularly this underserved, thin- file/unscorable market, they need to have data. If we can bring utilities and telecom data online, this could fundamentally change the way lending is done, for example, in urban America."

Making it work

The next step, says Turner, is the active phase-which means reaching out to the telecom and utility companies to show them the evidence that supplying data to credit bureaus helps their bottom line as well. The growth of lending based on nontraditional data holds promise for the millions of individuals who do not currently have credit scores, many of whom are low- and moderateincome individuals, reports still another study, Market Interest in Alternative Data Sources and Credit Scoring, issued in December 2006 by the Center for Financial Services Innovation (CFSI), Chicago. "As this type of lending grows, these consumers can benefit from better, cheaper access to capital to invest in businesses and homes. Assisting consumers newly abl\e to access credit to use the opportunity to build assets, for example, through long-term homeownership, will be one of the challenges of broader credit availability through use of alternative data," the CFSI report states.

The CFSI made a number of other key points:

* Before alternative data are to be used to make massmarket lending and insurance decisions, more work needs to be done, including identifying the most useful data points.

* Although mortgage lenders have been using alternative data for many years, scale will not be achieved until alternative data are included in automated underwriting processes and standards are developed for the secondary markets.

* Data providers and lenders must target the right market segments, better identifying those consumers for whom lending can be done more effectively based on the use of alternative data to assess creditworthiness.

All's fair

In August 2004, Fair Isaac unveiled its FICO Expansion(TM) product, which tapped nontraditional sources of consumer data not found at the national credit-reporting agencies in order to assess the credit risk of adults with minimal or no credit history on file. The concept from the beginning was that by using FICO Expansion scores for these consumers, including recent immigrants and young adults, businesses could make more financial services available. FICO Expansion uses the same score range (300-850) and similar scaling as classic FICO scores, but it was designed to score traditional credit bureau no-hits and thin-files-that is, to provide a risk-assessment tool when FICO scores are unavailable. Without being specific, Fair Isaac's Nelson elaborates on the types of consumer data used to create FICO Expansion scoring. "One data aggregator collects information about consumer debit accounts- if a consumer opened a new checking account or bounced a check. Other data providers provide similar kinds of positive and negative information about consumer's behavior in regard to phone bills or payment plan performance-i.e., whether the consumer always pays his magazine subscriptions or rent-to-own furniture bills," she says. All of these kinds of third-party, alternative or public-records data can be "very predictive" in trying to understand the risk of repaying debt, says Nelson. When FICO Expansion was launched in 2004, it was immediately tested, including usage by a couple of large lenders. In summer 2006, the company launched a new study enlisting a large group of top-tier lenders across different verticals, including mortgage lending (Freddie Mac; HSBC Finance Corporation, Prospect Heights, Illinois; and Option One Mortgage Corporation, Irvine, California), and using information provided by a new network of data providers. After looking at the results of its study, in November Fair Isaac claimed FICO Expansion had proven to be the first "strong and reliable credit score for assessing the risk of millions of Americans who have little or no credit information on file at Equifax, Experian and TransUnion." In the mortgage field, when Fair Isaac pulled in applications where there were no hits at the credit bureaus, it was able to find enough alternative credit bases to score 50 percent to 70 percent of those same applicants. When it came to borrowers with thin files, with the alternative credit bases, 65 percent to 85 percent now could be scored. "We were able to find that 65 percent of the consumers could score 640 or higher, which addresses the point that there are definitely creditworthy consumers who don't have credit bureau data," says Nelson. "Many underserved consumers are good credit risks for lenders."

All for one, one for all

In September 2006, Barrett Burns took on a new position as president and chief executive of VantageScore Solutions LLC, a Stamford, Connecticut, company launched by the three national credit- repository companies in March of the same year. Burns, who had worked at a number of financial institutions over his 30-year career, decided to take on one more challenge, and it is a unique one. The consumer credit information business is dominated by the big- three companies-Equifax, Experian and TransUnion. But the numerous finance and lending institutions that dealt with the repositories were a little frustrated because each one, when looking at the same information, could end up with a different score for the same borrower. "There are about 4.5 billion pieces of credit data that are reported every month in the three bureaus, and each uses different algorithms and weightings, which produce different scores," explains Burns. "Large lenders said to the bureaus, 'We need a better product.'" VantageScore^sup sm^ marked the first time the three companies joined forces to produce a model that scores consumers consistently across all three companies. The model itself was developed from an identical national sample of 15 million anonymous consumer credit files from each credit-reporting company. VantageScore was created as a result of client demand, says Joe Greenwald, senior director of marketing for Experian. "It was important to have a score that works across all three major credit bureau reporting companies. VantageScore is the first score that uses the same algorithm for each of the three bureaus, and it is very predictive [as to which consumers will go delinquent or not]. Predictability and consistency of scores have been two of the major attributes that our clients have been asking us for in development." There was a third reason as well. "The major lenders are well- saturated at the prime and near-prime segments of the market, so a lot of them were pushing the bureaus to address thin files," says Burns.

Tapping into the thin-file market

VantageScore addresses the thin-file sector in a number of ways. First, the company shows 15 different scorecards on its Web site, two of which are specific to thin-files. More important, VantageScore is, in a sense, more inclusive than the credit-score models used by the three individual credit bureaus. There are two ways that has been accomplished. "The primary reason we can score more of the population is due to a difference in what we call 'score-exclusion criteria,'" says Burns. "In other words, certain consumers are not scored. They have been excluded because they don't meet the minimum information criteria. For example, most products don't score customers who have not had at least one account open for six months or who have not had a trade updated in the last six months." What VantageScore does is extend the time criteria from six months to two years. In addition, the standard approach in credit scoring focuses on the worst performance in any account to determine whether that consumer is creditworthy, but VantageScore instead randomly selects one trade from a consumer credit report to determine whether that consumer is a good or bad risk. This is especially relevant for the mortgage industry, suggests Chet Wiermanski, vice president of TransUnion Analytics Decision Services, Chicago. "Under the traditional approach, credit bureaus look at the worst of a consumer's file, where he or she might have gone delinquent on a bank card, student loan or department-store credit card. However, there is a hierarchy as to how people repay their obligations; the last thing they will do is go delinquent on a mortgage." By using the random-tradeline approach, VantageScore can identify consumers who may have been delinquent on a bank card but always paid the mortgage. As a result, a person who may have been deemed a bad credit risk under the worst-on-file methodology is determined to be a good credit risk using the random-tradeline approach. "We were able to relax standards," says Wiermanski, "and by relaxing standards we were able to score more people." Adds Experian's Greenwald, "VantageScore is more forgiving than other scores. We are making an attempt to score as many consumers as possible, and that is in everyone's best interest. It helps consumers and the lenders, because it gives them a better chance to make a loan."

Specialization

The issue of thin-files and no-files in regard to certain minority groups has translated into low homeownership rates, maintains Leonardo Simpser, managing director of the Hispanic National Mortgage Association (HNMA), San Diego. In fact, his company was organized to do something about that problem, at least with regard to the Latino population. In November 2006, HNMA unveiled its Hispanic Automated Underwriting System (HAUS), which it says is the first automated mortgage underwriting system calibrated to make credit decisions using nontraditional characteristics prevalent in Latino and other immigrant communities. The idea behind HAUS is that it will enable mainstream lenders to expand market share to other population sectors and to sell these loans automatically to HNMA's new correspondent lending company. (HNMA and Frankfurt, Germany-based Deutsche Bank AG established a correspondent lending company, San Diego-based HNMA Funding Co., which has a $500 million warehouse facility to buy closed and HAUS- originated loans.) With all that, says Simpser, HAUS offers a turnkey, automated solution, giving Latinos and other immigrant borrowers access to customized loan programs and additional liquidity. "The reason we built this is to be able to offer credit to more people, to have a better way of looking at people who are not performing to credit-industry standards, and determine whether or not they are good credit risks," Simpser says. To build the model for HAUS, HNMA interviewed lenders to see how they handled the Latino and immigrant markets. Then HNMA scrutinized additional data to see whether that information was predictive. "We don't use FICO," Simpser notes. "Willingness to repay [a mortgage] is usually determined by FICO, and most of the people we are going after do not have a credit history-or if they do, it is not enoug\h for a FICO score. So, how do you create a credit history? It has to be done from nontraditional tradelines." Simpser wouldn't elaborate as to which nontraditional tradelines are being used in the HAUS model, other than to say it is designed to incorporate nontraditional credit and can accommodate information from multiple borrowers, borrowers with multiple income sources (including cash income) and borrowers with multiple employers, all of which are characteristics common within the Latino and immigrant communities. So far, Simpser says, HNMA has trained about a dozen customers, mostly smaller banks, that have committed to doing loans with the new company.

Reality

For a long time, people would say there was no relationship between income and credit scores or between race and credit scores- but that was not true, observes the Brookings Institution's Fellowes. "There is a very strong association between credit scores and income and race, but it is hard to explain what that relationship is because there are so many different variables involved," he says. Credit-scoring systems basically ignored the inherent social weaknesses in the programs because reports were, metaphorically, in black and white, indicating that a particular consumer was either bad or good. The mortgage industry, among other financial services sectors, has been looking for shades of gray in the reporting. If a consumer had 30 trades (uses of credit) and two were bad, maybe that person is creditworthy for a certain amount of money at a certain price, or that person is still good for a mortgage because mortgages are generally paid first even if someone gets behind on other bills. The overriding social issue, says Fellowes, is that weak consumer credit reports and low scores play a growing role in the ability of families to get ahead, influencing prices for loans, insurance, mortgages and even renting apartments. So, any progress made on this front ultimately may help some deserving underserved borrowers grow their personal wealth, as well as benefit lenders seeking to capture more emerging-markets business. According to Fair Isaac's data, 15 percent of the country's adult population falls into the thin-file category, while 10 percent would garner no "hits" whatsoever. The issue of thin-files and no-files in regard to certain minority groups has translated into low homeownership rates, maintains Leonardo Simpser.

Steve Bergsman is a freelance writer based in Mesa, Arizona. He can be reached at smbcomm@hotmail.com.