Metropolitan Policy Program The Brookings Institution Credit Scores, Reports, and Getting Ahead in America Matt Fellowes “The expansion of Findings Consumer credit reports and scores play a growing role in the ability of families to get consumer credit ahead, now influencing prices for loans and insurance and efforts to get jobs and rent apartments. An analysis of a quarterly sample of 25 million anonymous consumer credit reports and scores for every U.S. county between 1999 and 2004 reveals that: has created a ■ Consumer credit scores widely vary across counties, with the South having the highest concentration of consumers with weak credit scores. In 2004, among all con- windfall of bene- sumers, the average score on a credit score index maintained by one of the major bureaus was 656, out of a scale that ranges from 350 to over 850. Meanwhile, the aver- age credit score in the South was 635, and more than one in five borrowers in a typical fits to both Southern county have scores that suggest they are very risky borrowers.

■ Between 1999 and 2004, most counties with weak consumer credit scores saw economies and declines in the average consumer credit score, while counties with strong scores generally experienced modest gains. Nationwide, credit scores only modestly fell dur- ing this period, but the average Southern county experienced a larger decrease. consumers, but ■ Counties with relatively high proportions of racial and ethnic minorities are more likely to have lower average credit scores. This evidence does not suggest that a bias that growth has exists, or that there is a causal relationship between race and credit scores, raising questions for future research.

paralleled a sub- ■ High homeownership rates and county per capita income are strongly associated with high consumer credit scores. The average county with a low, mean credit score had a per capita income of $26,636 and a homeownership rate of 63 percent in 2000. stantial increase Meanwhile, the typical county with high average credit scores had higher per capita incomes ($40,941) and higher shares of homeowners (73 percent).

in the financial ■ Financial insecurity, primarily measured by the frequency of loan delinquencies, rose between 1999 and 2004. Over those five years, the proportion of mortgage bor- rowers 60 or more days late in their mortgage payments increased by 108 percent, from insecurity of one out of every 106 borrowers to one out of every 51. About one out of every 21 bor- rowers had at least one credit-bearing account 60 or more days past due in 2004.

Americans.” Consumer credit reports and scores are playing a growing role in the economic mobility of consumers today. But rising consumer debt and loan delinquencies mandate that govern- ment leaders, with their private sector partners, pursue a series of reforms to increase consumer education and responsibility, market accountability, and accuracy.

MAY 2006 • THE BROOKINGS INSTITUTION • SURVEY SERIES 1 Introduction Figure 1. Consumer Debt Has Substantially veryday, more than 27,000 Increased Over Time employees in the credit bureau industry walk into Total Consumer Debt (2005 dollars) over 1,000 locations around $2,500 Ethe country and process over 66 mil- 1 lion items of information. Out of this $2,000 massive churning of activity, credit bureaus produce consumer credit $1,500 reports and scores, two of the most powerful determinants of modern $1,000 American consumer life.2 (in billions) Among their many applications, $500 credit reports and scores now help determine if a family can borrow 0 money to buy major necessities like homes and cars; they affect the prices 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 businesses charge for such products as mortgages and auto insurance; reports and scores are used by an increasing Consumer Debt as a Percentage of Personal Income number of employers to assess job 25% applicants; they are used by landlords to evaluate prospective renters; and a 20% growing number of utilities are using credit reports and scores to price 15% deposits for numerous services.3 In short, both the access and terms of 10% access to an increasing array of basic necessities, including jobs, housing, 5% insurance, energy, and communica- tions, are now influenced by an 0 individual’s consumer credit report 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 and scores.4 Among other effects, the growing availability of credit has contributed to Source: Author’s analysis of data from the Bureau of Economic Analysis, Bureau of Labor Statistics, the recent surge in consumer debt and the Federal Reserve (Figure 1). Businesses have substan- tially expanded access to loan products among consumers as a result of the with individually tailored products. every 1,000 individuals (Figure 2).7 At capacity credit reports and scores give Pre-approved credit card offers, for the same time, bankruptcy is affecting businesses to predict lending risks.5 instance, are made possible by the a broader cross section of Americans, Among other benefits, this has readily available information in con- becoming, some experts say, a “middle increased access to assets like houses, sumer credit reports.6 class phenomenon.”8 given consumers more choices about While this expansion of consumer Among the numerous factors con- market products, and spurred eco- debt has created a windfall of benefits tributing to this growing financial nomic development in neighborhoods to both economies and consumers, its insecurity, consumers frequently point once ignored by creditors and insur- growth has paralleled a substantial to the proliferation of credit offers and ance companies. increase in the financial insecurity of to their inability to understand all of At the same time, consumer credit American consumers. As one indica- the many choices they are now con- information has provided businesses tion of this, between 1980 and 2004 fronted with in the financial services with a very sophisticated marketing the personal bankruptcy rate skyrock- market.9 Not understanding how to tool, creating countless new opportu- eted from about one out of every responsibly manage debt has become nities to target consumer segments 1,000 individuals to about five out of increasingly costly for families as debt

2 MAY 2006 • THE BROOKINGS INSTITUTION • SURVEY SERIES lions of borrowers, and drives-up the Figure 2. Personal Bankruptcy Filings Have Substantially cost of credit associated with those Increased Over Time scores.10 But, empirical evidence that speaks to that point, and the general Total Number of Personal Bankruptcy Filings distributional effects of nontraditional data on consumer credit scores, is 1.8 thin.11 There are also very different 1.6 assessments about the accuracy of 1.4 information currently collected by the 12 1.2 bureaus. 1 Also there are important questions left unanswered about the appropri- 0.8 ateness of market responses to credit (in millions) 0.6 scores. There is no public data avail- 0.4 able, for instance, that speaks to the 0.2 optimal level of mortgage price fluctu- 0 ation across different levels of risk. Put differently, the price-point where 1980 1981 1982 1983 1984 1985 1986 1986 1986 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 higher prices for mortgage borrowers with low credit scores becomes price- Bankruptcy Rate gouging rather than just cost-covering 6 is not clear. In a market that functions 5 perfectly, competition would drive prices to that point. But evidence that 4 consumers are under-informed cer- tainly creates an incentive for 3 over-charging.13 On the other hand, it is not clear that government can (or 2 should) set such a price point. 1 In short, the amount of information about credit reports, scores, and the 0 market applications of both, is out of (Number of Bankruptcies per 1,000 people)

1980 1981 1982 1983 1984 1985 1986 1986 1986 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 step with the importance these market products now play in the lives of con- sumers. Source: Author’s analysis of data from the American Bankruptcy Institute To begin to address the need for this information, this paper analyzes how information in credit reports, and one of the credit scores that information is has become more widely available. about their role overseeing the accu- used to calculate, varies across the Expanding uses of consumer credit racy of their credit report is now more country. We also examine the relation- reports and scores have consequently important than ever. ship between those scores and spurred benefits and costs to American More generally, the lack of visibility delinquencies and socioeconomic fac- consumers and local economies. But, of these issues has meant that a tors. To do this, we use a unique the geographic distribution of these research agenda geared toward under- database based on a sample of 25 mil- effects is not widely known, stunting standing credit reports and the ways lion consumer reports and scores in the development of appropriate, local they are used to set prices and inform every quarter between 1999 and 2004. policy responses to these market prod- market decisions is underdeveloped This information is used to build pro- ucts. Many consumers likely need and under-funded. On the most ele- files of consumers in every county in more information about credit report mental issue about the accuracy of the United States. and scores, for instance, as they grap- credit reports, for instance, there is a We also develop a policy agenda for ple with the increasing number of widely held sense that the bill payment political and business leaders to choices made possible by credit information not collected by bureaus respond to the fundamental role that reports. Similarly, consumer awareness drives down the credit scores of mil- credit reports and scores now play in

MAY 2006 • THE BROOKINGS INSTITUTION • SURVEY SERIES 3 the lives of American consumers and measure variables. For these reasons, Among the numerous uses of this the economy. It is time for government the numbers reported in this analysis information, bureaus create credit leaders, with their private sector part- should be interpreted as estimates. scores, or sell this information to other ners, to pursue a series of reforms to Besides the national estimates, we institutions that calculate their own increase consumer education and also report estimates for every county credit scores. Credit scores are calcu- responsibility, market accountability, in the country. For instance, we report lated differently for different market and accuracy. the average consumer credit score in a applications and for different compa- county, the proportion of consumers in nies. In general, though, scores are a a county that are in different credit function of numerous factors related Methodology score ranges, and the proportion of to the financial life of a consumer, borrowers in a county who are delin- including an individual’s payment his- About the Data quent on different types of loans. tory, debt-to-equity ratio, length of Data for this project was obtained These county level data allow us to credit history, extant types of extended from TransUnion’s trend database. look at how credit scores and informa- credit, and numerous additional vari- TransUnion is one of the three major tion in credit reports varies across the ables related to recent transactions.19 credit bureaus that collect financial country. At the same time, these data Although the weights assigned to each information on nearly every consumer can be linked with socioeconomic of these general classes of variables in the U.S. that has some type of information from counties, which are available for some credit scores, credit account. All available data in gives us the opportunity to analyze the specific variables and the specific the trend data were aggregated from how information in credit reports weight assigned to these variables depersonalized consumer credit varies with different county socioeco- remains the private property of the reports.14 nomic profiles. Also, county level data institutions that develop them.20 For this analysis, we use an anony- is just the type of local data policymak- Typically, consumer credit scores mous sample of American consumers ers need to understand the importance are scaled for ease of use to range that had a credit report on file with of credit scores in the lives of their between 350 and 850, where higher TransUnion between 1999 and 2004. constituents.17 numbers represent lower levels of In each of the 24 quarters during risks—or a lower probability of a delin- these six years, a random sample of About Credit Reports and Scores quency or default—and lower approximately 25 million borrowers Credit reports contain four general numbers indicate a higher level of risk. was extracted from the population of types of information. First, there is Nonetheless, each bureau, along with American borrowers with a credit information related to the identity of a the company that pools together the report. This sample was used to create consumer, including a consumer’s bureau’s reports into a single score national, quarterly estimates related to name, address, social security number, (myFICO), has a modestly unique credit scores and delinquencies in and date of birth, among other similar range. The TransUnion Auto Model every quarter between 1999 and information. Second, credit reports ranges between values of 300-900, for 2004.15 contain information related to an indi- instance. For this analysis, we rely on Importantly, some estimates in vidual’s use of credit-based products. TransUnion’s credit score in the trend TransUnion’s trend database differ For most consumers, this includes database, which ranges between 350 from more readily available estimates. information related to mortgages, credit and over 850. For instance, John M. Barron and his cards, retail credit cards, and auto colleagues report that 1998 Tran- loans, among others. A growing number About the Analysis sUnion estimates of bankcard of consumers also have their utility pay- To account for why consumer credit delinquencies differ from estimates ment histories reported and chronicled scores vary so widely across the coun- available from the American Bankers by credit bureaus, along with nontradi- try, we first consider how this Association, even if the metrics are not tional loans, like payday loans.18 Third, information varies across the major identical.16 Similarly, our own 2004 there is an inquiry history of applica- Census regions and 10 divisions. New estimates of bankcard debt differ from tions for credit. Finally, public record England consists of counties in Maine, other, recent estimates. To date, a full information related to a consumer’s New Hampshire, Vermont, Massachu- account explaining those differences is financial health is tracked by credit setts, Rhode Island, and Connecticut; not available, although the bureau’s bureaus. This includes declarations of the Middle Atlantic includes counties much larger population of reporting bankruptcy, along with a range of other in New York, Pennsylvania, New Jer- institutions may play a role in explain- related records that are publicly avail- sey; the South Atlantic includes the ing these differences. There are also able. Together, this information District of Columbia and counties in differences in the metrics used to comprises a consumer’s credit report. Delaware, Maryland, West Virginia,

4 MAY 2006 • THE BROOKINGS INSTITUTION • SURVEY SERIES Figure 3. Credit Scores Widely Vary Across U.S. Counties

Average Consumer Credit Score (2004) Lower Risk Area (Average credit score greater than 690)

Source: Author’s analysis of data in TransUnion’s trend database Higher Risk Area (Average credit score less than 624) Note: All available data in the trend database were aggregated from depersonalized consumer credit reports. Data are displayed by country and in quintiles.

Virginia, North Carolina, South Car- counties within these areas. It would cial insecurity across the country. We olina, Georgia, and Florida; the East have been more ideal to weight these concentrate on delinquencies as a South Central division includes coun- averages by the number of borrowers measure of financial insecurity, includ- ties in Kentucky, Tennessee, Alabama, in each county, but that information ing the proportion of consumers in a and Mississippi; the West South Cen- was not available. We assume that the county with any loan that is more than tral division includes counties in distribution of people across counties 60 days past due, along with the pro- Texas, Oklahoma, Arkansas, and closely resembles the overall distribu- portion of consumers that are Louisiana; the East North Central tion of borrowers across counties. delinquent on their mortgages and includes counties in Ohio, Michigan, Data for county populations is bankcards. All three measures are Indiana, Illinois, and Wisconsin; the based on the U.S. Census Bureau compiled by the same credit bureau West North Central includes counties county population projections. These that collected the consumer credit in Minnesota, Iowa, Missouri, North estimates are based on a method scores analyzed in this paper. Dakota, South Dakota; the Mountain referred to as the “administrative We also assess the relationship division includes Montana, Wyoming, records component of population between numerous socioeconomic Colorado, Utah, Idaho, Nevada, Ari- change” method. This technique characteristics and both consumer zona, and New Mexico; and the Pacific essentially uses a wide range of annu- credit scores and financial insecurity. division includes Washington, Oregon, ally available administrative records to The major characteristics we consider California, Alaska, and Hawaii. make inferences about population are: median county income, the pro- Because we have county-level data, change over time.21 portion of black county residents, the we make inferences about consumers Next, we consider how consumer proportion of Hispanic residents, and in these regions and divisions by tak- credit scores vary across counties to the proportion of homeowners in a ing population-weighted averages of determine the different levels of finan- county. The most recent data on these

MAY 2006 • THE BROOKINGS INSTITUTION • SURVEY SERIES 5 Figure 4. Southern Counties Have High Proportions of Consumers with Very Weak Credit Scores

Proportion of Low Consumer Credit Scores to All Scores (2004) Low (Less than 13.00%)

Source: Author’s analysis of data in TransUnion’s trend database High (More than 24.14%) Note: All available data in the trend database were aggregated from depersonalized consumer credit reports. Data are displayed by country and in quintiles; low consumer credit scores are less than 492, bottom 10 percent range of scores in the validation sample. characteristics for the entire popula- revolving loans are one-time lines of between 600 and 800 on this scale.23 tion of U.S. counties is the long form credit, such as automobile or educa- On either side of this central tendency, survey included in the 2000 U.S. Cen- tion loans, that usually close once the about 20 percent of the population sus.22 This form was completed by principal and any interest is paid off; has scores less than 600, and about approximately one out of every six and mortgage loans are all loans another 25 percent have scores above American households in 2000, or secured by a home. 800. While most consumers in the about 19 million different housing United States are clustered around the units. Using this information, Census average score, the “high” and “low” weighted the response to approximate Findings risk consumers are concentrated the population in each county of the together in fairly systematic ways country, creating the richest resource A. Consumer credit scores widely across the country. currently available to analyze socioeco- vary across counties, with the In general, consumers in the typical nomic characteristics across counties. South having the highest concen- Southern county have much lower Finally, this study includes an analy- tration of consumers with weak credit scores than elsewhere in the sis of specific types of loans in a credit credit scores. country. While the typical consumer in report, which includes revolving loans, Out of a possible range between 350 the Western, Midwest, and Northeast non-revolving loans, and mortgages. and over 850 (where higher numbers counties had credit scores that ranged Revolving loans, such as those avail- indicate lower risks for underwriting) between 660 and 675, the average able through credit cards or home the average consumer credit score was among Southern counties was 635 in equity loans, are lines of credit that 656 in 2004. Around this mean, there 2004. Because this scale is constant provide a continuous source of credit is a fairly flat distribution: about 55 across the country, this indicates that within some predetermined limit; non- percent of the population has scores the average consumer in a Southern

6 MAY 2006 • THE BROOKINGS INSTITUTION • SURVEY SERIES portions of consumers in these cate- Figure 5. Credit Scores are Strongly Related to Mortgage Terms gories represent the universe of all borrowers in each county in the coun- Annual Cost of a $150,000, 30 year, fixed rate mortgage $14,856 10% try. To assess counties with unusually Interest Rate $13,884 9% high and low proportions of borrowers $11,832 8% on either end of this credit score dis- $11,184 $10,680 $10,440 7% $10,236 $9,984 tribution, we look at the bottom and 6% top two intervals. In particular, the 5% proportion of a county’s residents with 4% very weak credit scores is measured as

3% the proportion of borrowers that have

2% scores below 492, which includes the

1% lowest two categories tracked by our

0% data source. The proportion of a 500–559 560–619 620–639 640–659 660–679 680–699 700–759 760–850 county’s residents with very strong (Credit Score Range) credit scores is measured as the pro- portion of borrowers that have scores Source: Author’s analysis of data from Fair Isaac. above 840, which includes the two highest categories tracked by our data source.26 county appears more of a credit risk in the past. Their credit report indi- Among the Southern counties in than the average consumer in other cates that they are a higher risk than our analysis, an average of over 22 areas of the country. This carries other consumers, and financial institu- percent of borrowers in a county has important implications for the cost of tions now can rationally pass on that very weak credit scores, or scores credit across different areas of the higher risk through higher prices, below 492. That suggests more than country. whereas in the past there may not one out of every five borrowers in a To see this, Figure 5 illustrates the have been an offer of credit extended. typical Southern county may not have relationship between consumer credit Still, those higher costs are not with- access to additional credit or, at the scores, interest rates, and the annual out costs that both consumers and very least, pays a substantial premium cost of a $150,000 mortgage.24 These communities should take seriously. for it. In contrast, about 16 percent of consumer credit scores are average Most importantly, higher prices take current borrowers in a typical county FICO scores, based on the three money off the table for other types of located in other major regions had FICO-branded scores estimated by investments that can help families get scores under 492 in 2004. each of the three major bureaus. ahead. Similarly, when high-risk bor- Why do Southern counties have According to the company, FICO rowers are clustered together, the such large proportions of borrowers scores are the “credit scores most effects of these higher costs may spill with low credit scores? lenders use to estimate risk.”25 The over into the community by draining Most clearly, the information in company illustrates on its web page consumer spending away from retail, these borrowers’ credit reports tends to the relationship between interest rates homeownership, home improvements, show more risky behavior, because of and credit score categories—a practice and educations. For both families and higher delinquency rates, higher debt also used by creditors. Here, interest leaders, it is therefore important to to equity loads, fewer lines of open rates range from a low of 5.3 percent understand consumer credit scores, credit, and so on. But these data do charged to consumers with scores and the strategies needed to improve not explain what is driving this more between 760–850 to a high of 9.3 per- scores. risky behavior. Why are Southern bor- cent charged to consumers with scores Credit report data do point to clus- rowers, for instance, more likely to fall between 500 and 559. That difference ters of extremely high risk and behind on payments than borrowers in adds up to nearly $5,000 every year in extremely low risk areas of the country other regions? Is it something to do extra payments that are charged to (Figure 4). To illustrate this we con- with systematic differences in their liv- consumers with scores in the lower sider the eight credit score intervals ing expenses or wages? Or, does it range. reported in TransUnion’s trend data- have something to do with the types of Certainly a high-risk borrower may base: scores less than 421, between businesses selling products in this be more than happy to pay an extra 422–492, 493–594, 595–700, region? That such large proportions of $5,000 to qualify for a mortgage that 701–795, 796–839, 840–850, and borrowers in this region stand out for they may have been turned down for greater then 850. Together, the pro- having such weak credit scores begs

MAY 2006 • THE BROOKINGS INSTITUTION • SURVEY SERIES 7 further research in the future. With- out that attention, large proportions of Figure 6. Counties with Strong Credit Scores Tend to Improve consumers in Southern counties will Over Time; Counties with Weak Scores Decline likely continue to have credit scores that limit access to credit products or only qualify them for very high-priced mortgages, auto loans, and all of the other credit-based products families rely on today. On the other hand, none of the U.S. regions stand out as having an unusu- ally high number of consumers with extremely strong credit scores, or scores in the highest two categories tracked by TransUnion’s trend data- base. Among all U.S. counties in 2004, an average of 12 percent of consumers in a county had scores above 840. Across the four major regions, the average county in each region with consumers in this extremely low risk Source: Author’s analysis of data in TransUnion’s trend database category varied between 11 and 14 per- Note: All available data in the TransUnion trend database was aggregated from depersonalized cent of all consumers, suggesting that consumer credit reports the typical county across each of the regions had about the same proportion of extremely low-risk consumers. 17 points by 2004, or from 611 in consumer credit scores are somewhat 1999 to 594 by 2004. path dependent, with particularly seri- B. Over time, most counties with But counties with average con- ous consequences for Southern weak consumer credit scores saw sumer credit scores in 1999 showed consumers. Areas of the country with sharp declines in the average nearly no discernable trend during higher average lending risks among consumer credit score, while this same period. Moreover, counties consumers saw that average lending counties with strong scores gener- that had very high average consumer risk increase over time, whereas areas ally experienced modest gains.27 credit scores in 1999 had modestly with lower average lending risks saw This trend is illustrated in Figure 6. higher values by 2004. In particular, that risk drop even lower over time. The average consumer credit score in counties that had higher average This trend points to a potentially 1999 for every U.S. county is listed on credit scores than 90 percent of the ruinous fiscal cycle for consumers the horizontal axis; the vertical axis other counties in the country in 1999 with low credit scores, recently exam- displays the absolute change in the saw that average modestly increase by ined in more detail by Dean S. Karlen average consumer credit score two points by 2004, or from 708 in and Jonathan Zinman.29 While credit between 1999 and 2004, by county. 1999 to 710 in 2004. scores create an opportunity to under- Organizing the data in this way illus- Some of the sharpest declines in write high-risk consumers that trates how consumer credit scores average, county credit scores occurred opportunity comes with a price, such change over time in counties with low, in the South. Between 1999 and as higher-priced mortgage loans. In average, and higher consumer credit 2004, the average credit score in turn, higher prices may make high- scores, relative to the rest of the coun- Southern counties fell by an average risk consumers more likely to miss bill try. The trend is unmistakable. of 10 points, compared to a nation- payments than consumers with strong Counties that had low consumer wide decrease of just two points. Even credit scores. Gaining access to credit scores in 1999 relative to the more dramatically, in both 1999 and higher priced credit, in other words, national average generally saw that 2004, 93 percent of the counties with may not always be a wise financial average drop in value by 2004. In par- average credit scores lower than 90 decision, particularly for families who ticular, counties with average credit percent of the other counties in the have trouble paying bills on time. This scores lower than 90 percent of the country were located in the South. risk is likely pronounced when lower other counties in the country in 1999 At least when looked at from this income households are the high-risk saw that average drop by an average of aggregated level, these trends suggest borrowers.

8 MAY 2006 • THE BROOKINGS INSTITUTION • SURVEY SERIES This trend may be reinforced by some questionable business practices Table 1. Race and County Credit Score Profiles used for high-risk borrowers. In the mortgage industry, for instance, some The average racial profile of counties in each risk category lenders sell negatively amortizing lines Proportion Proportion of credit to high-risk clients, most of Credit Score Risk Categories Black Hispanic which really should not buy this prod- 850–720 (Very Low Risk) 1% 4% uct. Negative amortization on a line of 700–719 5% 5% credit means that the monthly pay- 675–699 5% 8% ment for that credit does not account 620–674 12% 14% for the full amount of monthly inter- 560–619 28% 19% est charged for the credit. This may 500–559 (Very High Risk) 26% 23% make sense for high net-worth indi- viduals, or for individuals who only Source: Author’s analysis of data from TransUnion’s trend database and the U.S. Census Bureau. temporarily need to make small pay- Note: All available data in the TransUnion trend database was aggregated from depersonalized con- ments. But, for high-risk clients who sumer credit reports. Credit score categories correspond with the credit score risk categories reported can only afford a loan by accumulat- by Fair Issac in October 2005. ing more debt on their principal balance, this may lock their credit score into a downward path. Along those same lines, universal default policies on credit cards also Table 2. Wealth, Homeownership Rates, and County Credit may contribute to this trend. A univer- Score Profiles sal default policy means that a credit Mean Terms for card company automatically boosts a 30 year fixed-rate, the APR (annual percentage rate) on $150,000 an individual’s credit card if their County Wealth mortgage credit report shows that the borrower Annual has recently missed a payment on Credit Score Per-Capita Homeowner Unemployment Interest Payments another line of credit. This policy Risk Categories Income Rate Rate Rate Due means that one late payment can add 850–720 (Very Low Risk) $41,384 75% 4% 5.7% $10,404 up to higher payments on numerous 700–719 $40,946 73% 4% 5.8% $10,536 lines of credit, increasing the chances 675–699 $36,280 69% 5% 6.3% $11,160 for further missed or late payments. 620–674 $29,391 65% 6% 7.5% $12,552 In this way, high-risk borrowers may 560–619 $26,636 63% 7% 8.5% $13,884 spiral into lower and lower credit 500–559 (Very High Risk) $17,136 72% 10% 9.3% $14,856 score categories. Meanwhile, it is important to point Source: Author’s analysis data in TransUnion’s trend database and the U.S. Census Bureau. out again that credit scores have made Note: All available data in the TransUnion trend database was aggregated from depersonalized con- underwriting many high-risk con- sumer credit reports. Credit score categories correspond with the credit score risk categories reported sumers possible, where in the past it by Fair Issac in October 2005. may not have been even considered.

C. Counties with relatively high proportions of racial and ethnic against a borrower, a problem that the a substantial improvement over the minorities are more likely to have lending and insurance industry has often subjective measures used to lower average credit scores. grappled with in the past.30 The scores evaluate risk in the past. Still, we do Credit scores are an assessment about also do not include any information find that that county racial profiles are the level of numerous types of risk about the neighborhood that the bor- associated with the average credit posed by consumers, such as to a cred- rower lives in, which makes it difficult score of a consumer. In other words, itor or an employer. The models that to see how scores can redline commu- the higher the concentration of racial calculate credit scores do not include nities, another problem that has or ethnic minorities in a county, the a person’s race, so it is unclear how plagued these industries. more likely the county’s average credit these scores could discriminate For these reasons, credit scores are score will be low. This does not suggest

MAY 2006 • THE BROOKINGS INSTITUTION • SURVEY SERIES 9 that a bias exists, or that there is a credit, and numerous additional vari- In the 552 counties with a very low causal relationship between race and ables related to recent transactions. typical score, the per capita income is credit scores. Instead, this association Once these variables are accounted $26,636, about 6.7 percent of the reflects the numerous, historical dis- for, there may not be an independent labor force is unemployed, and about parities between races in the access to effect associated with a county racial 63 percent of the households own the and availability of high quality educa- profile.31 homes they live in. On the other tion, well-paying jobs, and access to Along those lines, there is evidence hand, the counties that have higher loans, among other factors. But the that the depth of an individual’s typical credit scores tend to have presence of this relationship does raise knowledge about credit scores, and the much more wealth and higher important questions that should be significance of these scores in a fam- employment rates. In particular, in explored through further research, ily’s financial life, is strongly related to the 270 counties with average credit particularly in instances where infor- many socioeconomic characteristics.32 scores between 700–719, the per mation in reports are being used in Less information about this product capita income is $40,946, or 54 per- nontraditional, under-researched mar- may lead to lower scores.33 Similarly, cent higher than the counties with ket applications, like screening job more risky credit behavior may be very low average credit scores. About applicants and pricing insurance. driven by systematic differences 4.2 percent of the labor force is To illustrate this relationship, first between borrowers with thin and full unemployed, which is quite a bit consider the simple associations credit files, the effects of unscrupu- lower than the national average. And, between a county’s racial and con- lous lenders of credit, and systematic about 73 percent of the households sumer credit score profile in Table 1. differences in the labor market. own the homes they live in, again Here, we use the credit score cate- There also may be an additional, higher than the national average. gories automatically generated by the albeit far more technical, force at work This evidence points to the strong, bureau to assess this relationship. here obscuring the true relationship underlining association between the Take the counties on either side of between race and credit scores. A typical consumer credit profile of a the central tendency: the approxi- recent analysis by economists at the community and the per capita income mately 552 U.S. counties that have Federal Reserve looked at problems and homeownership rates in a com- very low average consumer credit associated with selective reporting of munity. Counties with good consumer scores between 560–619, and the 270 account information to the bureaus. credit scores have more homeowners counties that have very high average Their analysis suggests that the use of and more personal income. scores between 700–719. In the coun- a borrower’s highest balance on a Why is this so? A big reason here is ties with a very low typical score, revolving line of credit as a substitute that credit scores act as a gatekeeper about 19 percent of the population is for the credit limit (which affects the for creditors, insurers, even potential Latino and another 28 percent is credit utilization component used to employers, in deciding if they should black. On the other hand, the counties calculate credit scores) may systemati- extend an offer to an individual. In that have higher typical credit scores cally drive down credit scores for those counties with lower average credit tend to be essentially all white coun- who “are at the margins of credit wor- scores, a higher proportion of people ties. In particular, in the 270 counties thiness.”34 Traditionally, that margin may be turned down for an application with average credit scores between has been occupied by more vulnerable for a loan or insurance than in a 700–719 only about 5.1 percent of the sectors of the economy. county with a higher proportion of population is Latino and just 1.1 per- people with higher average credit cent is black. D. High homeownership rates scores.35 When added up across all of This evidence does not suggest that and county per capita income are the consumers in a county, we see racial differences between counties strongly associated with high con- major differences in the comparative cause these differences in scores. sumer credit scores. homeownership rates of communities. Instead, these data point to an associa- Homeownership rates and county per In the example above, for instance, the tion, which frankly is not very well capita income increase as the typical communities with a very low average understood, and requires more rigor- consumer credit score in a county credit score have a homeownership ous analysis. For instance, the strong increases in value. These findings rate 16 percent lower than the com- association in Table 1 could be over- point to an important opportunity munities with a very high average stated because we have not accounted leaders have to have families get ahead credit score. for the variables factored into credit by boosting their creditworthiness. Differences in credit scores also scores, like an individual’s payment Take the counties on either side of have important implications for the history, debt-to-equity ratio, length of the central tendency that we exam- prices consumers are charged for credit history, extant types of extended ined in the previous section (Table 2). these necessities. In the counties with

10 MAY 2006 • THE BROOKINGS INSTITUTION • SURVEY SERIES very high average scores, for instance, the annual payments for a $150,000 Figure 7. By Design, Credit Scores are Strongly related mortgage would be about $10,536, to Delinquincies while the payments for the same loan 20% in the counties with lower average 18% scores would be $14,856 every year, a 16% 41 percent price premium. These higher prices make homeownership 14% possible for many high risk consumers. 12% But, that considerable benefit must be weighed against the effects of taking 10% money off the table for other types of 8% investments in getting ahead. A con- sumer may very well be better off, for 6% instance, investing that money and 4% repairing their credit scores before county by Borrowers, of Past-Due Proportion going ahead and buying a higher- 2% priced mortgage. 0 Together, these findings point to an 525 550 575 600 625 650 675 700 725 750 important opportunity leaders have to Average Credit Score, by county help families get ahead. Helping fam- Source: Author’s analysis of data in TrendUnion’s trend database ilies boost their creditworthiness in Note: All available data in the TrendUnion database was aggregated from depersonalized consumer communities with low credit scores credit reports would give more families that oppor- tunity to invest in assets, like houses. It also would likely lower the prices delinquencies increase in frequency. By design, falling behind on pay- they pay for a number of basic neces- This then has an influence on the ments has a substantial impact on sities, including financial and price for loans and insurance, and consumer credit scores. To illustrate insurance products. In turn, this consumer efforts to get jobs and rent these effects, Figure 7 charts the rela- frees up money for savings, retail pur- apartments. Since information about tionship between the proportion of chases, or other investments, like loan delinquencies is collected by borrowers with delinquencies in a education or retirement. credit bureaus, it provides a very rich county and the average credit score in To do this, leaders will need to glimpse at the financial security of a county. In a nearly linear manner, address the rising rate of delinquen- American across the country, along credit scores decrease in value as the cies, which we discuss in the next with very detailed information about incidence of delinquencies increases section. By design, this has a system- the performance of different types of in a county.36 This carries important atic effect on credit scores. But to loans, including mortgages and revolv- implications for the price of many boost credit scores, leaders also will ing lines of credit. The unmistakable necessities along with access to an need to look more closely at the rea- impression from these data is that even larger number of necessities. sons why consumers are deemed to be financial insecurity has increased in We consider these effects by analyz- a high risk by creditors, insurers, recent years, though the South and ing either end of the delinquency employers, utility companies, and any particular lines of credit are largely distribution, or counties with the of the various other institutions that driving this trend. highest and lowest proportion of bor- utilize these data. rowers with a delinquent loan. At the Overall Financial Insecurity low end, this includes the counties E. Financial insecurity, primarily The ratio of borrowers who are late that have a higher proportion of delin- measured by the frequency of paying their credit-bearing accounts quent borrowers than 90 percent of loan delinquencies, rose between has increased in recent years. In 1999 the other counties in the country, a 1999 and 2004 about one out of every 23 borrowers group we refer to as counties with The status of payments made on out- was over 60 or more days behind in very high delinquency rates.37 At the standing loans is a critical variable their payments. In 2004, that number high end, this includes the counties factored directly into a credit score. In had grown to about one out of every that have a lower proportion of delin- a nearly linear manner, consumer 21 borrowers, about a 9 percent quent borrowers than 90 percent of credit scores decrease in value when increase. the other counties in the countries, a

MAY 2006 • THE BROOKINGS INSTITUTION • SURVEY SERIES 11 Figure 8. Southern Counties Have Higher Consumer Delinquency Rates Than Other Areas of the Country

Ratio of Past-Due Borrowers to All Borrowers (2004) Low (Less than 3.30%)

Source: Author’s analysis of data in TransUnion’s trend database High (More than 6.96%) Note: All available data in the trend database were aggregated from depersonalized consumer credit reports. Data are displayed by country and in quintiles; past-due borrowers are 60+ days past due. group we refer to as having very low development opportunity.38 growth in consumer delinquency rates delinquency rates. Just as average credit scores widely out of any other census division, In the counties with very high delin- vary between U.S. counties, delin- growing by nearly 23 percent between quency rates, the average credit score quency rates, and the effects they hold 1999 and 2004. The next fastest in 2004 was 597. In contrast, among for consumers and local economies, growth rates during this time period the counties with very low delinquency also systematically vary across the were in Midwestern counties, which rates, the average credit score in 2004 country (Figure 8). Standing out again saw the average delinquency rate grow was 702. One of the important conse- from the other regions of the country, by about 15 percent during this quences of these differences is the Southern counties have a higher pro- period. But the average delinquency higher prices charged for many neces- portion of delinquent borrowers than rate is much lower in the Midwest sities when credit scores are low. borrowers in other regions of the than the national average, suggesting When average county credit scores are country. In 2004, nearly six out of that there is a lower overall level of low the average price of many of these every 100 borrowers in the typical financial insecurity, even if it is necessities is likely high. Capturing a Southern county had a delinquent increasing at a faster pace than the portion of those extra costs in commu- loan, compared to about five borrow- rest of the country. nities with severe delinquencies could ers out of every 100 in every other Meanwhile, counties in the West create extra spending power among region of the country. actually saw nearly no change in consumers in these areas. To the Even more telling, delinquency delinquency rates between 1999 and extent that these delinquencies are rates have been rising in the South 2004, increasing by just about two behavioral, lowering delinquency rates faster than in other regions of the percent during this period—a trend among consumers might very well rep- country. Counties in the Southeast largely driven by counties in the resent a significant economic states experienced the fastest recent Pacific. Counties in the Northeast

12 MAY 2006 • THE BROOKINGS INSTITUTION • SURVEY SERIES area of the country also saw more seven out of every 10 households growing number of American home- modest increases in delinquency rates owned at least one home, up by about owners are finding the costs of during this period compared to the eight percent from 1992. Not only is homeownership too financially bur- national average, increasing by about this the highest homeownership rate densome, throwing in jeopardy what six percent. recorded since the Census Bureau for most families is their primary asset. These clear regional differences in began tracking this statistic in 1965, But, as with the overall delinquency delinquency rates illustrate an impor- the recent jump in homeownership is rate, this trend is reflected unevenly tant, and widely overlooked, spatial also the largest sustained increase on across the country. dimension of financial insecurity in record.39 Once again, consumers in Southern the United States. While rates of But, the promise of homeownership counties face a risk that is substan- delinquencies have increased nearly has turned into too large of a financial tially larger than consumers who live everywhere in recent years, counties in burden for a rapidly growing number elsewhere in the country, even though the South stand out as having the of homeowners (Figure 9). This is the housing stock in the South is the swiftest increases in delinquency rates reflected by the large number of least expensive in the country. Not in the country. homeowners that are delinquent on only are there higher proportions of What explains these higher delin- their mortgage payments. On this homeowners with delinquent mort- quency rates in the South? It’s count, it is important to point out that gages in this region, their ranks have certainly true that the South tends to there are many different sources of also swelled at a clip unmatched by have a much lower median income data on mortgage delinquencies, from nearly any other area of the country. than other areas of the country, which each of the three credit bureaus, to Leading the pack, counties in the may mean that consumers have less surveys of financial establishments like Alabama, Kentucky, Mississippi, and money available to make payments that administered by the Mortgage Tennessee Southeast division saw the than consumers elsewhere. But hous- Bankers Association, to surveys of bor- average mortgage delinquency rate in ing prices and other major costs of rowers like the Survey of Consumer a county nearly triple between 1999 living tend to be much lower in the Finances. The estimates across all these and 2004, including about one of South than in other areas of the coun- different sources are often different, and every 32 mortgage borrowers by 2004. try, which should depress the it is still not clear why this is the case. In stark contrast, the average West- significance of the earnings gap. At the Although our estimates are based on ern Pacific county actually saw only a same time, the unemployment rate in an enormous sample of consumers, it 14 percent increase in the number of the South is lower than in both the should be treated as an estimate, just delinquent mortgages during that time Midwest and the West. For these rea- like all of these other estimates.40 period. So, while most counties saw a sons, it’s not at all clear that basic Our data indicate that about one surge in the number of borrowers economic differences between the out every 50 homeowners with an unable to pay their mortgages on time, regions explain the particularly high active mortgage was delinquent on there are clearly regions of the country delinquency rates in the South. Future their mortgage payments in 2004. that have emerged as leaders and lag- research will need to explore this vex- That certainly does not mean that gards in this phenomenon. Although ing issue. each of these borrowers will lose their home prices have generally surged well homes. But, because of the effects ahead of income increases in recent Home Loan Insecurity credit scores have on several major years, it is actually where homes are Overall measures of financial insecu- necessities, the costs of living for these least expensive that we find the high- rity in a county paint a broad picture homeowners will likely increase, est incidences of homeowners falling of growing financial insecurity through which may make them more likely to behind in payments. out the country. Still, this overall lose their homes in the future. impression masks the underlining Meanwhile, the rate that mortgage Revolving Debt Insecurity financial commitments made by con- delinquencies increased in recent Much has recently been written about sumers to create this impression. One years has substantially outpaced the surge in revolving debt, including of the most important of these under- increases in overall delinquency rates. debt held in credit cards and home lining components is the growing Between 1999 and 2004, the propor- equity loans.41 This is certainly true. propensity of homeowners to fall tion of homeowners behind on their According to our data, the total real behind in their mortgage payments. mortgage payments jumped 115 per- value of revolving debt held per bor- Throughout the 1990s, homeowner- cent, increasing from one out of every rower was valued at about $8,900 in ship surged across the country, adding 100 mortgage borrowers to about one 2004, a 46 increase from 1992.42 a substantial asset to the lives of mil- out of 50. This remarkable leap in But, the rate of delinquencies on lions of Americans. By 2004, about mortgage delinquencies means that a revolving loan accounts has actually

MAY 2006 • THE BROOKINGS INSTITUTION • SURVEY SERIES 13 Figure 9. Southern Counties Have Higher Mortgage Delinquency Rates Than Other Areas of the Country

Ratio of Past-Due Mortgage Borrowers to all Mortgage Borrowers (2004) Low (Less than 1.30%)

Source: Author’s analysis of data in TransUnion’s trend database High (More than 3.43%) Notes: All available data in the trend database were aggregated from depersonalized consumer credit reports. Data are displayed by country and in quintiles; delinquent mortgages are 60+ days past due. decreased during this period. Accord- tain division of the country have the So, even while the proportion of ing to TransUnion’s trend database, an highest rates of delinquency rates on delinquent borrowers with revolving average of one out of every 40 borrow- revolving loans in the country, includ- lines of credit remain generally less ers with a revolving line of credit was ing more than one out of every 33 than found in the mortgage market, delinquent on their payments in 2004, borrowers with this line of credit. many more people own revolving lines a 10 percent decrease from 1999. Right behind these counties, delin- of credit than mortgages, meaning that Once again, though, these trends quency rates on revolving lines of the insecurity associated with these played out unevenly across the country credit include about one out of every late payments, including the effects on (Figure 8). 31 borrowers in the South Atlantic credit scores, affects many more peo- Proportions of borrowers with counties that carry this form of debt. ple. At the same time, the smaller revolving lines of credit in the Pacific This evidence again speaks to the value of payments typically made to division of the country saw the spatial implications of these national maintain revolving lines of credit may sharpest drops in the rates of delin- trends. The South, home to some of mean that borrowers have an easier quencies between 1999 and 2004, the least expensive places to live in the time meeting these payments than falling by about 19 percent. On the country, also is home to some of the much higher mortgage payments. other hand, delinquency rates in four highest proportions of borrowers with Census divisions—New England, revolving debt who can not meet their South Atlantic, West North Central, payments on time. Because the effect and the Southeast—saw delinquency these late payments have on credit rates on revolving lines of credit fall by scores, consumers in these areas will two percent or less during this period. end up paying higher prices for a num- Meanwhile, the counties in the Moun- ber of major necessities.

14 MAY 2006 • THE BROOKINGS INSTITUTION • SURVEY SERIES Research and Policy low credit scores. respondents with a low educational Implications We also found that some of the fac- attainment, and 64 percent of respon- tors that credit scores have replaced dents with a low income, indicated redit reports and scores from underwriting decisions—like that they knew that their credit rating affect both the access and race—are strongly associated with weakened when they missed a credit terms of access to a large credit scores. Some of this may have card payment. and growing array of basic to do with the fact that the depth of an Even while there is limited public Cnecessities, including jobs, housing, individual’s knowledge about credit understanding of credit reports and energy, and communications. Besides scores, and the significance of these scores, there has never been more formally separating issues of class, scores in a family’s financial life, is information about both. Hundreds of race, and gender from underwriting strongly related to many socioeco- credit counseling agencies now pro- decisions, information in credit reports nomic characteristics.44 vide information about credit reports is also used to substantially expand the What implications do these findings and scores, consumer finance TV per- range of market products available to raise for policymakers? Most impor- sonalities constantly advertise consumers. Among other benefits, this tantly, leaders must take more information, and there are thousands has increased access to assets like seriously the complexity of participat- of documents posted on the Internet houses, given consumers more choices ing in the modern American about both.47 Lack of information, about market products, and spurred marketplace. The need for a sustained then, is not driving this low public economic development in neighbor- financial education has never been understanding. hoods once ignored by creditors and greater. Policy and business leaders Instead, there is likely a low under- insurance companies. At the same also need to continue asking if these standing of credit reports, and the time, consumer credit reports and assessments of consumers are always significance they play in people’s lives, scores have spurred a dramatic surge accurate, while also assessing if the because people are not accessing this in the availability of consumer debt, many market responses to these risk information. Unless an individual is in which has improved the lifestyle of assessments are reasonable. We financial trouble, there are no routine countless Americans.43 expand on all of these implications ways that people are exposed to credit With these benefits, however, there below: reports, and their many applications. have come some costs. Consumers are This absolutely vital ingredient to a confronted with an array of choices Provide a Financial Education for person’s ability to get ahead is really that they do not fully understand. Consumers only recognized after someone is in Applications of these products to mar- Although credit reports and scores financial trouble. This makes credit ket decisions are not well researched. play a fundamental role in the finan- reports both an asset that goes unuti- And, consumers have amassed a sub- cial life of individuals, few people lized, while at the same serving as an stantial amount of debt. As one understand that role. How many peo- unnecessary roadblock for others. possible outcome, we found the delin- ple know, for instance, that credit The solution to building awareness quencies across a range of different reports affect the access and terms of and knowledge of credit reports lies, lines of credit substantially increased access to financial services, insurance, then, in promoting routine ways that in recent years. In turn, growing finan- telecommunications, apartments, and give consumers access to this informa- cial insecurity drives down credit even jobs? tion. For this reason, leaders should scores, which affects the access and In fact, a recent survey by Providian implement financial education for terms of access to a broad array of Financial and the Consumer Federa- consumers. This can be incorporated basic necessities and undermines tion of America indicates that few into existing curriculum at the K-12 household goals for savings and wealth people understand credit scores, one level. Banks can integrate basic infor- accumulation. of the primary uses of credit reports.45 mation about credit reports when they Many of these trends vary in sys- Although 93 percent of consumers sign their clients up for service. They tematic ways across the country, even indicated that they knew credit scores could also make access to a variety of dividing the country into clear geo- are affected by missing payments, only budgeting tools part of their customer graphical areas that lead or lag the 27 percent knew what credit scores service operations. Consumer organi- nation. Most strikingly, counties in the actually measure. Moreover, the zations could provide feedback South are sharply distinguished from scarce knowledge about the impor- information about credit counseling the rest of the country by the high tance of credit scores that does exist agencies in local markets, and invest rates of delinquency in the typical systematically varies with an individ- in web-based financial education county, which drives the high propor- ual’s income and educational courses. Insurance companies can tion of consumers who have extremely attainment.46 Only 56 percent of the share with customers the impact their

MAY 2006 • THE BROOKINGS INSTITUTION • SURVEY SERIES 15 scores are having on their rates; insur- give consumers the option to receive a as we have discussed. There also may ance regulators could provide free credit score estimate from the have been beneficial effects associated information about the different ways bureaus or MyFICO, which makes it with non-traditional applications, like that companies they regulate use this more difficult to link reports with insurance pricing and landlord deci- information.48 And, utility and score predictions. All of this means sion-making. telecommunications companies can that the FACT act made some impor- Still, the rising delinquency rates inform their customers about the role tant steps forward, but more needs to make it important to now ask if credit scores play in their access to services. be done. For those households who reports and scores are always being Through all of these ways, consumers have not received news of this law, not used in a way that is beneficial to all can receive the information they need acted on it, or cannot afford to buy consumers. What costs, in other to become proficient in negotiating their credit score, it is uncertain how words, are created along with the sub- through this tool, or barrier, to getting their interest is being served by this stantial and widely-recognized benefits ahead. new policy. associated with these products, and To address this limitation, a CBDA how are those costs distributed across Pass a Credit Bureau Disclosure Act should require that bureaus annually both place and people? Policymakers should pass a Credit submit reports to federal policymakers To start with, there may be an over Bureau Disclosure Act (CBDA) that that speak to the accuracy of their reliance on credit scores if other infor- requires the credit bureaus, rather information. Most importantly, mation that speaks to the true risk a than policymakers, to report their bureaus should report the results of person represents is not considered. As assessments of the accuracy of their consumer inquires they receive into Peter McCorkell of Fair Isaac and information. The bureaus already the accuracy of their information. Company has noted, “ignoring other regularly assess the accuracy of their Such an act would preserve the rights relevant information in the mortgage model predictions because their of these businesses to maintain their decision process is not in the best business depends on it. But, theses intellectual property, while also safe- interests of either borrowers or assessments are not shared with poli- guarding the privacy of consumers. lenders….[But] during the mortgage cymakers, or the public. Just as important, it would create an and refinancing boom there was cer- Is this necessary? Various independ- incentive for a long overdue assess- tainly an economic motivation to move ent researchers have tried to assess the ment of how the accuracy of both on to the easy cases rather than spend accuracy of the information in the credit scores, and the reports they are extra time on the difficult ones.”50 In reports, along with the accuracy of the based on, can be improved. this case, credit scores are being relied risk estimates. But sample sizes are on too heavily because they are not often small, time bound, and based on Research market responses to credit the only factor that should be used to information from a single bureau. The report information assess risk. result: there is a lot of conflicting evi- This report has illustrated how prices Similarly, there is no rule or even a dence and uncertainty about the respond to changes in credit scores general guide that specifies what a accuracy of these market products. through an analysis of pricing deci- proper market response in any of these This fuels ad hominem opinions about sions made in the mortgage market. cases should be to credit scores. There the accuracy of these products, which When credit scores are low, mortgage is no fixed cap, for instance, that sug- is unfair to the bureaus, the busi- borrowers pay more—in our example, gests a broker cannot systematically nesses that rely on their data, and thousands of dollars more every year— add 2, 5, or even 10 percent to the consumers. for their loan than borrowers who have total value of a loan in fees and higher To address just this type of uncer- high credit scores. But, this example interest rates when lending to some- tainty, the Fair and Accurate Credit really just scratches the surface of how one with a low credit score. Aside from Transactions Act of 2003 (FACT Act) markets now respond to credit reports concerns addressed in the Equal required the bureaus to make one free and scores. We also know that prices Credit Opportunity Act, no govern- credit report available to consumers for and access to other loans, like ment policy specifies what is fair and every year, which gives consumers the home equity or auto loans, are what crosses the line. ability to more easily assess the accu- affected by credit scores. So are prices This raises a key, albeit very compli- racy of the information contained in and access to auto and home insur- cated, question for policymakers: the report.49 But, the law did not ance, along with access to jobs and What is the optimal level of price fluc- require the bureaus to report a) the apartments. tuation across different levels of risk? number of queries they receive or b) Oftentimes, credit scores have sub- In other words, at what point do information about the outcome of stantially expanded the opportunity of higher prices for borrowers with low those queries. The law also did not high-risk households to access credit, credit scores stop covering the costs of

16 MAY 2006 • THE BROOKINGS INSTITUTION • SURVEY SERIES that 35 percent of companies used credit reports the predicted higher level of risk, and Endnotes for this purpose, up from 19 percent in 1996. This start becoming price-gouging market closely paralleled findings in the 2003, University products? 1. Robert Hunt, “A Century of Credit Reporting in of Florida, National Retail Security Survey. C) In theory, an invisible cap exists in America.” (Federal Reserve Bank of Philadelphia, There are private companies which provide credit- the market that curbs the extent to Working Paper No 05-13, 2005); and U.S. Census checking services for landlords. For instance, see which prices vary across and within Bureau, County Business Patterns. http://www.amerusa-tenant-screening.com/. markets. Competition between busi- 2. Credit reports preceded credit scores. Reports were There is no evidence that we have been able to find nesses, in this ideal world, drives this originally gathered by local or regional bureaus to that attests to how many renter applications are cap to its most efficient point. But, we track the performance of loans made by individual accepted or rejected because of their credit scores, don’t live in an ideal world. Unfortu- creditors, from banks to retail establishments. Then although the existence of nationwide companies to provide this information suggests there is a large nately, it is currently difficult to know in 1958, Bill Fair and Earl Isaac developed a model for generating credit scorecards, which, by the end market for this information. D) On the basis of what an appropriate cap would look of the 1970s, were used by most banks. The draw- background conversations with state regulators, we like, or whether a cap is even a proper back of these early scores, however, was that they know that it is commonly believed that a growing policy response, without more informa- were often not comparable across either industries number of utility companies are using credit report tion. We know that markets are or businesses within industries, because scorecards information. But, the market impact of these appli- responding to these market products, were custom developed by bureaus for different cations is essentially unknown.

but we don’t always have a very good clients. In fact, it was really not until a 1995 rec- 4. An increasing number of insurance companies use sense of how they respond, particu- ommendation by Fannie Mae and Freddie Mac that “insurance scores”, which are constructed from larly in non-traditional applications of mortgage lenders use FICO scores in their mort- information in credit reports. Insurance scores are these products. For this reason, policy- gage lending decisions that the use of credit scores used to predict the likelihood of a future adverse makers need to begin systematically began to explode in this country, now influencing event, like an insurance claim. State policymakers gathering and evaluating information billions of decisions every year. For more informa- are increasingly acting to curb the use of credit tion about the evolution of this market product, see related to market responses to credit reports by insurance companies. Hawaii bans the Lyn C. Thomas, “A Survey of Credit and Behavioral use of credit scores, and all states regulate its use. A reports and scores. Public policy is Scoring: Forecasting Financial Risk of Lending to premature; research is not. growing number of statehouses are also considering Consumers.” International Journal of Forecasting, legislation to limit the legal uses of credit scores in As a model, federal policymakers 16 (2000):149–72; and Hollis Fishelson-Holstine, insurance pricing, such as justifying an increase in should consider the results from the “Credit Scoring’s Role in Increasing Homeowner- an insurance premium. In this case, state authority Illinois pilot predatory lending data- ship for Underserved Populations” in Nicolas P. to regulate insurance is in tension with provisions base, created in July 2005 (Public Act Retsinas and Eric S. Belsky, Building Assets, Build- in FACT that preempt state limits on credit report No. 94-280). This law authorizes the ing Credit (Cambridge: Joint Center for Housing uses. Studies and Washington: Brookings Institution Illinois Department of Financial and 5. In particular, the “high risk” market was mostly Press, 2005). Professional Regulation to develop and ignored prior to the invention of credit scores, maintain a database on mortgage 3. A) The permissible uses of credit reports were iden- which are based on information contained in credit products sold in Cook County over the tified in the Fair Credit Reporting Act (1970) to reports. The prediction power that these scores next four years. Consumer credit protect the privacy of individuals. However, the have given companies substantially expanded their scores, types of mortgage products, range of legal uses is quite broad. Among the many ability to extend credit. For more information, and the price of those products are legal applications, credit reports can be used for please refer to Nicolas P. Retsinas and Eric S. Bel- any purpose related to a credit transaction (includ- among the many variables that will be sky,. Building Assets, Building Credit: Creating ing marketing related products), underwriting Wealth in Low-Income Communities (Cambridge included in this database. This will be insurance, evaluating credit risks, and for employ- the first public database in the country and Washington: The Joint Center on Housing and ment purposes. New state laws/regulations The Brookings Institution, 2005); and Thomas A. to systematically collect the full of range somewhat curb these applicable uses, although Durkin and Michael E. Staten,. The Impact of Pub- of information analyzed by lenders these laws are uneven across the country. Also, the lic Policy on Consumer Credit (Boston: Kluwer when making access and pricing deci- 1996 amendments to FCRA somewhat curb the use Academic Publishers, 2004). sions.51 Among its many attributes, of reports, by allowing consumers to sue when their 6. Mark Furletti, “An Overview and History of Credit these data will provide policymakers report is used for a purpose not permitted by Reporting,” Federal Reserve Bank of Philadelphia, with the capacity to assess the extent FCRA. At the same time, the Fair and Accurate TransUnion Workshop (held April 5, 2002). to which mortgage prices respond to Credit Transactions Act (FACT), passed in Decem- changes in credit scores. Given the far ber 2003, preempted the states from passing 7. The Bankruptcy Abuse Prevention and Consumer certain laws that curb the content and permissible reach of credit reports and scores, it Protection Act of 2005 (signed into law in April uses of credit reports. B) There are numerous sur- 2005), may very well continue to reduce the num- seems quite reasonable to collect simi- veys that suggest employers are more frequently ber of personal bankruptcies (at the time of lar data to study other applications of using credit reports to screen job applicants. A publication, bankruptcy filings have substantially credit reports and scores. 2004 survey of companies by the Society for dropped). Among its several major provisions, the Human Resource Management, for instance, found law requires credit counseling for filers and makes

MAY 2006 • THE BROOKINGS INSTITUTION • SURVEY SERIES 17 it more difficult for consumers to file for Chapter 7 15. Consumer credit scores do vary across bureaus, paper related to this relationship rely on these data bankruptcies, which accounted for about 74 per- although the extent to which they vary is hardly cer- (www.myfico.com). cent of all personal filings in 2004. Chapter 7 tain. There is no evidence that aggregate level 25. MyFICO homepage [www.myfico.com] bankruptcy filings are more popular than Chapter data—the type of data we analyze here—signifi- 13 filings because outstanding debt is cancelled cantly varies from place to place. For this reason, 26. Although there is no hard set rule, it is frequently after non-exempt assets are sold-off. On the other we don’t have a reason to suspect that reliance on reported that scores above 650 will qualify a bor- hand, Chapter 13 filings require filers to establish a the evaluations of one bureau for county level sta- rower for a prime mortgage rate. Consumers with five-year pre-payment plan. tistics should systematically bias our analysis. scores above 840 may not be offered different terms than consumers with scores of 800. We select this 8. Teresa Sullivan, Elizabeth Warren, and Jay 16. Barron, Elliehausen, and Staten. 2000. . cutoff point only because TransUnion’s trend data- Lawrence Westbrook, The Fragile Middle Class: 17. The Fair and Accurate Credit Transaction Act of base has this value as a cutoff point. Americans in Debt (New Haven: Yale University 2003 mandated that the Federal Reserve and the Press, 2000). 27. To determine whether the selected time period Federal Trade Commission explore these relation- affected this analysis, we also looked at the percent- 9. See for instance Timothy Egan, “Newly Bankrupt ships and report back to Congress. The first of age change between 1999 and 2003, 1999 and Raking in Piles of Credit Offers,” New York Times, those reports was due in December 2005, but, at 2002, and 1999 and 2001. All of these analyses December 11, 2005, p. 1. the time of publication, that report had not yet confirm the trend illustrated in Figure 5. been released. 10. “It is estimated that more than 70 million Ameri- 28. Upcoming analyses by the Federal Reserve will exam- cans make rent, mortgage, and other recurring bill 18. The Brookings Institution’s Urban Market Initiative ine panel data, which will yield insight into how this payments that are not reported to traditional credit hosted a roundtable on alternative data sources in trend plays out among individual borrowers. bureaus. As a result these consumers often have credit scoring, and has made available most of the lower credit scores than they should, and pay more presentations on their webpage: http://www.brook- 29. Dean S. Karlan and Jonathan Zinman, “Observing for housing, credit, and insurance than they ings.edu/metro/umi/20051215_paidroundtable.h Unobservables: Identifying Information Asymme- deserve.” Terry Clemans, Executive Director of tm. tries with a Consumer Credit Field Experiment.” National Credit Reporting Association, Inc., quoted (Dartmouth University, 2005). 19. The importance of these different categories in the in an October 3, 2005 Press Release by NCRA, Inc. calculation of a credit score varies, but myFICO 30. Concerns about unequal access to credit across 11. Information Policy Institute, “Giving Underserved reports a weight distribution for the general popula- racial and income groups led to the creation of the Consumers Better Access to the Credit System: The tion: 35 percent payment history, 30 percent Home Mortgage Disclosure Act (HMDA) and, then, Promise of Non-Traditional Data” (2005). amount owed, 15 percent length of credit history, the Community Reinvestment Act (CRA). For a (http://www.infopolicy.org). 10 percent types of credit used, and 10 percent new recent summary of HMDA see Robert B. Avery, credit [www.myfico.com, accessed September Glenn B. Canner and Robert E. Cook,. “New Infor- 12. For some examples see: Consumer Federation of 2005]. Debt-to-equity ratios, for instance, are not mation Reported under HMDA and Its Application America and National Credit Reporting Associa- used in credit card lending, and are not often in Fair Lending Enforcement.” Federal Reserve Bul- tion, “Credit Score Accuracy and Implications for included in a score for mortgage lending (but as a letin, Summer (2005): 354–394. For a recent Consumers.” (Washington: Consumer Federation of separate consideration). assessment of the CRA see Michael Barr, “Credit America, 2002); Michael E. Staton and Fred H. Where It Counts: Maintaining a Strong Commu- Cate, “Accuracy in Credit Reporting,” in Nicolas P. 20. For instance, myFICO reports the general weights nity Reinvestment Act” (Washington: Brookings Retsinas and Eric S. Belsky (2005); Robert B. assigned to classes of variables on its webpage. The Institution, 2005). Avery, Paul S. Calem, and Glenn B. Canner (2004); same information is not available for other credit Alison Cassady and Edmund Mierzwinski, “Mis- scores, such as those calculated by the three major 31. We did run a regression analysis that predicted takes Do Happen: A Look at Errors in Consumer bureaus or by insurance companies. credit scores as a function of a variety of informa- Credit Reports.” (Washington: National Association tion found in credit reports along with county 21. For more information, please refer to the web page of State PIRGs, 2004). socioeconomic information. We found that there of the Population Division of the U.S. Census, was a significant, direct relationship between credit 13. For evidence of consumers being under-informed, http://www.census.gov/popest/counties. scores and median income, proportion black, pro- see the annual survey administered by Providian 22. When fully implemented, the American Commu- portion Latino, and educational attainment. But, an Financial and the Consumer Federation of America nity Survey (ACS) will become a superior source of analysis of the robustness of these results suggested or GAO, “Credit Reporting Literacy: Consumers information. that the variables included in a credit report have a Understood the Basics but Could Benefit from Tar- much more powerful effect on credit scores than geted Educational Efforts.” (GAO-05-223, 2005). 23. The same distribution is reported by myFICO. See: socioeconomic characteristics. Still, we do not Systematic overcharging would likely lead to MyFICO. “Understanding your Credit Score,” report these results, and only note them here, adverse selection among clients if it were applied available at www.myfico.com/Offers/ because the information available from the reports across the board. More likely, evidence of informa- myFICO_UYCS%20booklet.pdf. was not sufficient for the model to fully control for tion asymmetries among consumers may lead to 24. Current rates are August 2005. The range of con- information accounted for in a credit score. Future systematic effects that substantively vary across sumer credit scores in the private bureau data used analyses, based on all of this information, are still consumer segments. in this report is nearly identical to the Fair Isaac needed to sort out the direct, indirect, and total 14. For more information about these data, please refer credit score, or the FICO score. Fair Isaac updates relationship between credit scores and socioeco- to John M. Barron, Gregory Elliehausen, and a daily, national estimate of the relationship nomic variables. It is for this reason that we only Michael E. Staten, “Monitoring the Household between its credit scores and the average interest report associational relationships and are very Sector with Aggregate Credit Bureau Data.” Busi- rate charged by mortgage lenders for several major explicit that our research does not point to causal ness Economics (35) (2000): 63–76. mortgage products. All additional analyses in this claims in this case.

18 MAY 2006 • THE BROOKINGS INSTITUTION • SURVEY SERIES 32. See annual survey administered by Providian 39. These statistics come from the Current Population eds, The Impact of Public Policy on Consumer Financial and the Consumer Federation of Amer- Survey, available at Credit( Boston: Kluwer Academic ica; or GAO, “Credit Reporting Literacy: http://www.census.gov/hhes/www/housing/hvs/his- Publishers, 2004). Consumers Understood the Basics but Could Bene- toric/index.html (August, 2005). 51. The 2004 version of the Home Mortgage Disclo- fit from Targeted Educational Efforts” GAO-05-223 40. The trend data is a sample of borrowers, which sure Act data took a big step forward with the (2005). includes both solely owned and joint accounts. For inclusion of price information. But, it still does not 33. This premise that more information leads to better this reason, it may be the case that some estimates have information about credit scores. To be sure, scores was one of the motivations for passing the are biased in types of lines where joint accounts are this is a difficult undertaking, since so many differ- Fair and Accurate Credit Transactions Act of 2003 particularly prevalent. The extent of this bias ent credit scores exist. The Illinois effort here may (FACT Act). This required the bureaus to make one remains to be seen. very well provide a model for federal policymakers free credit report available to consumers every year, as they determine how to include these data. 41. For instance, see Stan Sienkiewicz,. “Credit Cards which gives consumers the ability to more easily and Payment Efficiency” (Federal Reserve Bank of assess the accuracy of the information contained in Philadelphia, Payment Cards Center Discussion the report. Paper, 2001); and Tamara Draut and Javier Silva,. 34. Robert B. Avery, Paul S. Calem, and Glenn B. Can- “Borrowing to Make Ends Meet: The Growth of ner, “Credit Information Reporting and the Credit Card Debt in the ‘90s” (Washington:: Practical Implications of Inaccurate or Missing Demos, 2003). Information in Underwriting Decisions,” BABC 04- 42. Do keep in mind that this is the average among all 11 (Harvard University: Joint Center for Housing consumers with revolving debt. This means that the Studies, 2004). effect of home equity loans on this average is 35. The average credit score of a county is not used to weighted down by the much larger number of con- assess the costs of credit. Rather, the higher propor- sumers with credit cards and retail cards, which tion of consumers with high or low credit adds up tend to have lower balances than home equity to systematic differences across counties. loans.

36. Credit scores are generally used by businesses to 43. But, see: Gregg Easterbrook,. The Progress Paradox: assess risk within markets, and the range within a How Life Gets Better While People Feel Worse (New market likely differs from a national range. Still, York: Random House, 2003).. consumer credit scores are generated from a com- 44. See annual survey administered by Providian mon set of criteria, which means that the depth of Financial and the Consumer Federation of Amer- risk posed in a particular market by a score should ica; or GAO,”Credit Reporting Literacy: Consumers only marginally vary across markets. At the same, it Understood the Basics but Could Benefit from Tar- is important to note that there are many different geted Educational Efforts.” (GAO-05-223, 2005).. types of credit scores in use in the market today. The three major national bureaus, for instance, 45. For a summary of this survey please refer to: each generate its own credit score and a FICO www.consumerfed.org/pdfs/Providian_Press_Rel score, which is directly comparable across the three ease_9_05.pdf (October 2005). Also, see the GAO major bureaus. Although these bureaus account for report mentioned in notes 44 and 46. over half of the market, there are over 1,000 differ- 46. This finding was also reported in a recent analysis ent credit bureaus, many of whom generated some of credit report literacy by the Government type of predictive score. Accountability Office, GAO-05-411SP, which was 37. The overall delinquency rate in a county is the pro- mandated in the 2003 FACT Act. portion of borrowers in a county who have at least 47. Robert M. Hunt,”Whither Consumer Credit Coun- one loan that is 60 days or more past due. Mortgage seling?” Business Review, Quarter 4 (2005). delinquency rates are the proportion of borrowers 48. According to the Michigan Department of Insur- with at least one mortgage 60 days or more past ance, some companies price insurance by using due; revolving loans are the proportion of borrowers multiple “insurance score” categories, whereas with at least one revolving loan 60 days or more other companies used just two categories and oth- past due. Delinquency is measured by bureaus as ers do not use credit information. Drivers may be any account that is past due, but the data are only able to find a better price for insurance if they had available for accounts 60 days or more past due. this information. For this reason, our estimates of delinquency should be interpreted as very conservative esti- 49. To access your free credit report, go to mates. https://www.annualcreditreport.com/cra/ index.jsp 38. Delinquencies can also be brought on by unex- pected events, such as an unexpected health costs 50. Peter L. McCorkell, “The Impact of Credit Scoring or loss of a job. and Automated Underwriting on Credit Availabil- ity,” in Thomas A. Durkin and Michael E. Staten,

MAY 2006 • THE BROOKINGS INSTITUTION • SURVEY SERIES 19 Acknowledgments The Brookings Institution Metropolitan Policy Program thanks the Annie E. Casey Foundation for its generous support of our research on programs and policies that support lower-income working families, and the Fannie Mae Foundation, the John D. and Catherine T. MacArthur Foundation, the George Gund Foundation, and the Heinz Endowments for their general sup- port of the program.

The author is grateful to Evan DeCorte and Naotaka Sugawara for their research assistance. He would also like to thank a number of people who provided comments on earlier versions of this paper, including Ezra Becker of TransUnion, Alan Berube of Brookings, Glenn B. Canner of the Federal Reserve Board, Robert M. Hunt of the Federal Reserve Bank of Philadelphia, Bruce Katz of Brookings, Amy Liu of Brookings, Clifton O’Neal of Tran- sUnion, Sally Park of the National Foundation for Credit Counseling, Pari Sabety of Brookings, and Michael Turner of the Information Policy Institute. All errors remain the sole responsibility of the author.

For More Information: Matt Fellowes Senior Research Associate Brookings Institution Metropolitan Policy Program [email protected]

For General Information The Brookings Institution Metropolitan Policy Program (202)797-6139 www.brookings.edu/metro To receive weekly updates of Metro Program products, please subscribe at http://www.brookings.edu/metro/metro.htm

The Brookings Institution

1775 Massachusetts Avenue, NW • Washington D.C. 20036-2188 Tel: 202-797-6000 • Fax: 202-797-6004 www.brookings.edu

Direct: 202-797-6139 • Fax/direct: 202-797-2965 www.brookings.edu/metro