Analyzing Trends in Subprime Originations and Foreclosures: A Case Study of the Boston Metro Area

Analyzing Trends in Originations and : A Case Study of the Boston Metro Area Cambridge, MA Lexington, MA Hadley, MA Bethesda, MD Washington, DC Chicago, IL Cairo, Egypt Johannesburg, South Africa September 2000 Prepared for Neighborhood Reinvestment Corporation Abt Associates Inc. 55 Wheeler Street Cambridge, MA 02138 Prepared by Debbie Gruenstein Christopher E. Herbert

Analyzing Trends in Originations: A Case Study of the Boston Metro Area NEIGHBORHOOD REINVESTMENT CORPORATION, THE NEIGHBORWORKS NETWORK AND THE NEIGHBORWORKS CAMPAIGN FOR HOME OWNERSHIP 2002 This report was supported by the Neighborhood Reinvestment Corporation NeighborWorks Campaign for Home Ownership 2002. Neighborhood Reinvestment Corporation was established by an act of Congress in 1978 (Public Law 95-557). A primary objective of the Corporation is to increase the capacity of local communitybased organizations to revitalize their communities, particularly by expanding and improving housing opportunities. These local organizations, known as NeighborWorks organizations, are independent, resident-led, nonprofit partnerships that include business leaders and government officials. All together they make up the NeighborWorks network. The NeighborWorks Campaign for Home Ownership 2002 is the largest national initiative of its kind: a joint effort by private industry and government working with community-based NeighborWorks organizations to bring more families into home ownership. NeighborWorks organizations participating in the campaign use the NeighborWorks Full-Cycle SM system. Under this system, prepurchase education, innovative loan products and early-intervention delinquency counseling are combined into a system that helps create successful homebuyers who take charge of their neighborhoods as well as their homes.

Analyzing Trends in Originations: A Case Study of the Boston Metro Area Executive Summary This study uses available data to examine the growth of subprime lending in the Boston metropolitan area. Nationally, subprime lending for home mortgages has grown rapidly during the 1990s, with recent estimates indicating that this market segment accounted for about $200 billion in new loans by the end of the decade. On the positive side, the growth in subprime lending expands borrowing opportunities for those with impaired credit, and it may increase homeownership opportunities for those who might otherwise not qualify for a conventional or perhaps any mortgage. However, subprime loans carry higher interest rates and therefore create a greater payment burden for the borrower. And the rise in subprime lending has also been associated with a growth in predatory lending practices, in which lenders use misleading or fraudulent means to lure borrowers into loans with high interest rates and high fees that rob owners of their equity and threaten their ability to maintain ownership. lending is also much more prevalent in lowincome and minority communities, threatening the stability of these communities. As a result, this study analyzes not only subprime lending trends in the Boston metropolitan area overall, but also examines trends in neighborhoods of different income levels and racial composition. Data reported under the Home Mortgage Disclosure Act (HMDA) can be used to analyze trends in home loan origination and can provide valuable information on subprime lending trends at the neighborhood level. Analysis of HMDA data for the Boston area for the period 1994 to 1998 reveals that: Loan originations by subprime lenders grew by 435 percent between 1994 and 1998, compared to the growth of all loan originations by 119 percent. The share of all originations accounted for by subprime lenders more than doubled between 1994 and 1998, starting at less than 2 percent and climbing to 4 percent. The growth in subprime lending was much more significant for properties in low-income and minority neighborhoods than for properties in other kinds of neighborhoods. Between 1994 and 1998, subprime lending grew by 1,075 percent in very low-income neighborhoods (where the median household income is less than 50 percent of the metro area median) and by 827 percent in low-income areas (where median income is between 50 percent and 80 percent of the metro area median). In neighborhoods where more than half the residents are minority group members, subprime lending increased by over 1,000 percent. The market share of subprime lenders is also significantly higher in low-income and largely minority communities. By 1998, subprime lenders accounted for 10 percent of all originations in very low-income neighborhoods 2.5 times the subprime market share in the Boston area overall. In largely minority areas, subprime lenders accounted for 13 percent of all originations in 1998 more than 3 times their share for their entire metropolitan area. The concentration of subprime lending in minority and low-income communities was especially striking in the refinance market. September 2000 i

Analyzing Trends in Originations: A Case Study of the Boston Metro Area Analyzing Trends in by Lenders Data on home loan foreclosure sales listings from1995 through 1999 were obtained from the Boston Foreclosure Report, a local company that publishes monthly listings of properties to be sold at foreclosure auctions. The database includes the name of the lender pursuing each foreclosure, and we matched these names to the list of HMDA reporters to identify subprime lenders. An analysis of Boston PMSA foreclosures started by subprime lenders reveals: The overall volume of foreclosures declined by 30 percent between 1995 and 1999, but the volume of foreclosures started by subprime lenders grew by 154 percent. Among lenders reporting under HMDA, the share of foreclosures attributed to subprime lenders now exceeds their share of originations. Originations by subprime lenders peaked at 5 percent of all originations in 1997, while foreclosures by these lenders accounted for 11 percent of all foreclosures in 1999. And while the share of originations by subprime lenders declined in 1998, the share of foreclosures by these lenders was continuing to rise through 1999. The share of foreclosures by subprime lenders grew rapidly in Boston neighborhoods of all income levels and have come to account for between 10 and 12 percent of all foreclosures. by subprime lenders were also most common in majority minority neighborhoods. In 1999, the share of foreclosures in these areas attributable to subprime lenders was 14 percent, compared to about 10 percent in other areas. loans reached foreclosure much more quickly than prime loans. The median age of subprime loans at foreclosure was three years, while for other loans the median age was seven years. The relatively short period of time from origination to foreclosure of these loans may be an indication that these loans are not affordable for borrowers even at origination. September 2000 ii

Analyzing Trends in Originations: A Case Study of the Boston Metro Area Introduction One of the most striking features of home finance in the late 1990s was the rapid growth of the subprime lending. The term subprime generally refers to loans made to borrowers who have low credit ratings, but it also includes loans with very high loan-to-value loans. Prior to the 1990s, there were few borrowing options for homeowners with impaired credit. Fannie Mae and Freddie Mac (the government-sponsored secondary market enterprises or GSEs) would not purchase loans that did not meet their underwriting standards, and portfolio lenders were generally reluctant to take on the greater risks of such loans. But with the advent of securitization of all types of financial assets, mortgage bankers were able to tap the broader financial markets for funds to make loans that did not meet the conventional underwriting criteria of the GSEs or depositories. Estimates by the Mortgage Bankers Association indicate that subprime lending grew from $80 billion in 1993 to $210 billion in 1998, an increase of 162 percent in just five years. One of the characteristics of subprime loans is that they bear higher rates of interest than prime (or conventional) loans. The higher rate of interest is due both to the greater credit risk of these loans and the fact that most conventional loans benefit from favorable interest rates available through the GSEs. At the same time that subprime lending has been growing, the mortgage industry has also been developing more sophisticated methods for assessing the degree of credit risk associated with different loan and borrower characteristics. lending is thus also related to the development of risk-based pricing, in which borrowers theoretically will be charged the interest rate that is appropriate to the level of risk associated with their loan. There are both positive and negative aspects to the development of the subprime mortgage market and of risk-based pricing generally. On the one hand, these developments have increased options for borrowers who do not qualify for conventional loans. Many industry observers have touted the advent of risk-based pricing as a way of expanding homeownership opportunities for borrowers who otherwise would be excluded from the market. On the other hand, the ability to charge borrowers much higher interest rates introduces greater possibilities for unscrupulous lenders to take advantage of borrowers. The term predatory lending has been coined to describe the abusive practices used by some lenders to lure borrowers into loans that charge excessive interest rates and fees and leave borrowers with such unmanageable financial burdens that they are likely to end up losing their homes. 1 While not all subprime lending is predatory, subprime and predatory loans share the distinction of charging borrowers interest rates that are higher than conventional rates. Furthermore, even subprime loans that are not marked by excessive costs or deceptive practices will expose borrowers to higher risks than conventional loans due to the higher financial burden they entail. Thus, paradoxically, the higher interest rates that are imposed on the supposedly higher-risk recipients of subprime loans actually increase the risk that these borrowers will be unable to repay their loans. 1 For a thorough discussion of lender actions that may be defined as predatory see Deborah Goldstein, Understanding Predatory : Moving Toward a Common Definition and Workable Solutions, Neighborhood Reinvestment Corporation and the Joint Center for Housing Studies of Harvard University, October 1999. September 2000 1

Analyzing Trends in Originations: A Case Study of the Boston Metro Area Housing and community advocates have become increasingly alarmed as they have seen more and more borrowers lose their homes as a result of predatory lending practices. In many cases, the lenders appear to have targeted low-income and minority areas. 2 As a result, the foreclosures are not just wreaking havoc on the lives of individual borrowers, they may also be destabilizing low-income and minority communities. The present study relies on Home Mortgage Disclosure Act (HMDA) data on loan originations to provide empirical information on the following questions: What has been the trend in the volume of subprime home loan originations in the Boston area? How concentrated has subprime home loan lending been in low-income and minority neighborhoods? It is important to note that, while the concern about subprime lending is largely focused on predatory lending practices, it is not possible with the data available to distinguish between prudently employed risk-based pricing and predatory lending. However, based on a characterization of lenders principal lending activities, we can to some degree identify subprime loans. As a result, while this study is motivated by concerns about predatory lending, it focuses more generally on subprime lending. I. Trends in Originations A. Using HMDA Data to Track Our analysis of loan originations in Boston relies on data collected under the 1974 Home Mortgage Disclosure Act (HMDA). HMDA mandates that lenders disclose certain information on their home mortgage applications, including the race, census tract, and income of borrowers and the action taken on the application. However, not all financial institutions are required to report this information. HMDA compliance is mandatory only for lenders that meet certain annual loan volume and asset size requirements and originate loans in metropolitan areas. 3 In addition, HMDA reporting is not mandated for all types of loans. While lenders must disclose information for first mortgages, refinancing of first mortgages, and home improvement loans, second mortgages and home equity loans 4 are not subject to HMDA reporting requirements. loans may be first mortgages used either to purchase a home or to refinance an existing first mortgage (and possibly consolidate other consumer debts into the new first mortgage). But many subprime loans are second mortgages used to tap home equity for purposes other than home improvements. As a 2 For a discussion of the concentration of subprime lending in low-income and minority communities see Daniel Immergluck and Marti Wiles, Two Steps Back: The Dual Mortgage Market, Predatory, and the Undoing of Community Development. Woodstock Institute, Chicago, IL, November 1999. 3 For a thorough discussion of HMDA reporting requirements see A Guide to HMDA Reporting: Getting it Right!, Federal Financial Institutions Examination Council, 1998. 4 HMDA reporting of second mortgages and home equity loans are required if they are used for purchase or home improvement and are carried on lender s books as such. However, if the lender does not classify these loans as home improvement or purchase, they are not subject to HMDA reporting requirements. September 2000 2

Analyzing Trends in Originations: A Case Study of the Boston Metro Area result, it is likely that many subprime loans are not reported in HMDA. So, while HMDA is the best available information source on lending activity at a local level, it provides conservative estimates of subprime activity. Identifying subprime loans from HMDA data is not a straightforward task. One simple definition of a subprime loan is whether it meets the underwriting criteria used by the GSEs. 5 Another method of identifying a subprime loan would be based on the interest rate or fees charged. Unfortunately, HMDA does not contain the information needed to assess either of these factors. As a result, loans as reported under HMDA can be categorized as subprime based only on lender, rather than loan, characteristics. Using information from a variety of sources, the Department of Housing and Urban Development (HUD) has identified the lenders reporting under HMDA that specialize in subprime lending. This study uses this HUD list of subprime lenders reporting in HMDA to identify subprime loans. 6 This method appears to capture a significant portion of subprime loans reported in HMDA, but it certainly does not identify all subprime loans. Many primarily conventional lenders have entered the subprime market. The method used to identify subprime loans for this study will not encompass these subprime loans made by conventional lenders. B. Findings 1. Overall Trends in Boston Our analysis for the Boston Metropolitan Statistical Area (MSA) shows that subprime lenders more than doubled their market share over the course of four years and by 1998 accounted for four percent of all loan originations. Between 1994 and 1998, HMDA-reported loan originations in Boston increased by 119 percent (Exhibit 1), from around 82,000 to almost 180,000. At the same time, subprime originations increased from 1,356 to 7,250, an increase of 435 percent. 7 These disparate growth rates resulted in a rise of the subprime share of all originations, from less than 2 percent in 1994 to almost 5 percent in 1997. The share did decrease to 4 percent in 1998, in part due to the conventional refinancing boom that occurred during 1998. 5 A loan may not meet the GSEs underwriting criteria because it exceeds the GSE limit on the size of loans they can purchase. Of course, loans which exceed the GSE loan limits ( jumbo loans) may in all other respects be considered prime loans and so would not be counted as subprime. 6 For a thorough discussion of the methodology used by HUD to identify subprime lenders see Randall M. Scheessele, 1998 HMDA Highlights, Appendix D, U.S. Department of Housing and Development, Office of Policy Development and Research, Housing Finance Working Paper Series HF-009, 1999. In short, Scheessele developed annual lists first by using trade magazines and industry sources to identify subprime lenders. Next, he created a set of lender characteristics (such as high-denial rates, high refinance rates, large amounts of missing data, low percentage of FHA originations or GSE sales, etc) to screen HMDA data for possible subprime lenders. Finally, Scheessele attempted to speak with each of these potential subprime lenders to verify the characterization of their lending activity. Scheessele classified institutions as subprime if 50 percent or more of originations were subprime. 7 A somewhat higher rate of growth is to be expected for subprime lending, since the level subprime loans was much lower than the total lending volume in 1994. September 2000 3

Analyzing Trends in Originations: A Case Study of the Boston Metro Area Exhibit 1 Lenders' of HMDA Loans, 1994-1998 (Boston PMSA) Year 1994 81,841 1,356 1.7% 1995 66,272 1,147 1.7% 1996 91,424 2,732 3.0% 1997 93,684 4,533 4.8% 1998 179,486 7,250 4.0% % Change 119% 435% 144% Source: Abt Associates Inc. tabulations of HMDA data. 2. Trends in by Income Level Analysis at the neighborhood level shows that growth in subprime lending is strongest in low- and very low-income neighborhoods. In addition, subprime lenders share of all lending is highest and growing fastest in low-income areas. One of the concerns about the subprime market is that lenders have targeted their marketing efforts to low-income neighborhoods. 8 In order to determine lending activity within various types of communities, we combined HMDA data, which identifies the Census tract for each loan, with Census data on median income by tract. Loans were then divided into four categories based on median income: 9 High-income tracts are those with a median income greater than 120 percent of the median income of the entire Boston MSA; 10 Moderate-income tracts have median incomes between 80-120 percent of the MSA median income; Low-income tracts have median incomes between 50-80 percent of the MSA median; and Very low-income tracts have incomes less than 50 percent of MSA median income. One interesting feature of subprime lending trends was that growth in subprime loans was very high in each neighborhood income category (Exhibit 2). Between 1994 and 1998, the volume of subprime lending in high- and moderate-income neighborhoods increased by 188 percent and 488 percent respectively. But subprime originations in low-income neighborhoods grew even faster, with lending activity up by 827 percent in low-income tracts and 1,075 percent in very-low income tracts. 8 HMDA does report the borrower s income, so it would also be possible to analyze trends in subprime lending to borrowers of different income levels. However, since this research is concerned about neighborhoods, categorization at the tract level is more appropriate. 9 A small share (less than one percent) of HMDA loans could not be matched with income information, either due to missing tract information in HMDA or because the tract numbers that were listed in HMDA were not recognized by the Census database. 10 Based on a 1989 median income of $36,051 from the 1990 Decennial Census. September 2000 4

Analyzing Trends in Originations: A Case Study of the Boston Metro Area 1200% 1000% 800% 600% Exhibit 2 Growth in Loan Originations by Neighborhood Income Level 1994-98 (Boston PMSA) 1075% 827% 488% 400% 200% 188% 0% Very Low (<50%) Low (50-80%) Moderate (80-120%) Tract Median Household Income as Percent of Metro Area Median Source: Abt Associates Inc. analysis of HMDA data. Note: See Appendix 1 for loan volumes. High (>120%) In addition, the subprime growth rates were much higher than the total lending growth rates within each neighborhood income category. Consequently, the market share of subprime lenders grew in each type of neighborhood (Exhibit 3). The growth in market share was particularly high in lowand very low-income neighborhoods where the market shares of subprime lenders increased by 4 fold and 5 fold, respectively. lenders share of originations in very low-income tracts went from 2 percent to 12 percent in 1997 and then fell to 10 percent in 1998 (even though the level of subprime lending continued to grow rapidly in these neighborhoods between 1997 and 1998). The subprime market share in low-income neighborhoods started at 2 percent in 1994 and reached 8 percent by 1998. 14% 12% 10% 8% 6% 4% 2% Exhibit 3 of Originations by Lenders by Neighborhood Income Level (Boston PMSA) 0% Very Low (<50%) Low (50-80%) Moderate (80-120%) Tract Median Household Income as Percent of Metro Area Median 1994 1995 1996 1997 1998 High (>120%) Source: Abt Associates Inc. analysis of HMDA data. Note: See Appendix 1 for loan volumes September 2000 5

Analyzing Trends in Originations: A Case Study of the Boston Metro Area The concentration of subprime loans in low-income communities was especially striking in the refinance market. 11 In very low-income neighborhoods, the share of all refinance loans originated by subprime lenders grew from less than 4 percent in 1994 to over 25 percent in 1997. In low-income areas, the subprime share of refinancing increased from 3 percent to 14 percent over the same time period. And, although the boom in the conventional refinancing market resulted in a declining subprime share of all refinancing loans between 1997 and 1998, the growth in subprime refinance originations in very low-income and low-income communities was 39 percent and 98 percent, respectively, during that year. 12 3. Trends in, by Racial Composition As with income, both the growth rate of subprime originations and the subprime market share of all lending are highest in neighborhoods with high concentrations of minorities. The same methodology that allowed us to categorize loans into income groups allowed for classification by racial composition. 13 Based on census tract information on population by race, we created four categories of racial composition: high, moderately high, moderately low, and low. 14 High minority tracts are defined as any tract with minorities making up more than 50 percent of the population. The moderately-high minority category includes tracts with 20-50 percent minority population. The moderately-low and low categories are defined as between 10-20 percent minority and less than 10 percent minority population, respectively. Considering the high degree of correlation between income and race, it is not surprising that subprime lending growth rates and market shares appear to be highest in high-minority neighborhoods. While the level of subprime originations grew very strongly in each racial composition category, subprime lending activity grew by more than 493 percent in the moderately high minority tracts and by 1,005 percent in the high minority tracts (Exhibit 4). In addition, the market share of total loan originations comprised by subprime lenders grew faster in neighborhoods with high proportions of minority residents (Exhibit 5). While subprime lending remained between 2 percent and 5 percent of all lending in low and moderately-low minority communities between 1994 and 1998, the subprime share of lending in high-minority neighborhoods climbed from 2 percent in 1994 to 13 percent in 1998. During the same time, the subprime share of total lending in moderately-high minority neighborhoods increased from 2 percent to 6 percent. 11 The refinance market comprised between 35 and 44 percent of all HMDA-reported loan originations between 1994 and 1997, but jumped to 67 percent in 1998. See Appendix 2 for composition of loans by loan purpose. 12 See Appendix 3 for volumes of total and subprime refinance activity in metropolitan Boston. 13 A small share (less than one percent) of HMDA loans could not be matched with information on racial composition, due to missing tract information in HMDA. 14 Minorities are defined here as any non-whites, including whites of Hispanic origin. September 2000 6

Analyzing Trends in Originations: A Case Study of the Boston Metro Area Exhibit 4 Growth in Originations 1994-98 by Neighborhood Racial Composition (Boston PMSA) 1200% 1000% 1005% 800% 600% 400% 493% 462% 389% 200% 0% High (>50%) Moderately High (20-50%) Moderately Low (10-20%) Low (<10%) Source: Abt Associates Inc. analysis of HMDA data. Note: See Appendix 1 for loan volumes. Tract Percent Minority Exhibit 5 of Originations by Lenders by Neighborhood Racial Composition (Boston PMSA) 14% 12% 10% 8% 6% 4% 2% 0% High (>50%) Moderately High (20-50%) Moderately Low (10-20%) Low (<10%) Source: Abt Associates Inc. analysis of HMDA data. Note: See Appendix 1 for loan volumes. Tract Percent Minority 1994 1995 1996 1997 1998 September 2000 7

Analyzing Trends in Originations: A Case Study of the Boston Metro Area As was true in low-income neighborhoods, the concentration of subprime refinancing loans was especially high in areas with high proportions of minority residents. 15 In these largely-minority communities, the market share of subprime lenders in the refinance market grew from about 5 percent in 1994 to almost 26 percent in 1997. Again, the conventional refinancing boom decreased the subprime market share in these communities in 1998. Despite the fact that subprime refinancing grew by more than 73 percent between 1997 and 1998, the subprime share of all refinancing loans declined in 1998. However, in 1998 the subprime share of the refinance market was still more than 18 percent in these communities, compared to less than 5 percent in the metropolitan area overall. II. Trends in A. Using Data on in the Boston PMSA to Track The foreclosure analysis is based on data obtained from the Boston Foreclosure Report (BFR). Among the information contained in this database are the following pieces of information for foreclosed properties: Street address, town, county and state of foreclosed property Mortgagor name Mortgagee name Date that the property was listed in foreclosure report Year of the original mortgage There are several issues with this data that are worth noting. First, because the foreclosure report does not contain census tract information, we had to identify census tracts by geocoding properties based on the street address and town. Although the geocoding process was largely successful despite a lack of zip code information, about 9 percent of the observations could not be geocoded and, therefore, are absent from our analyses of foreclosures by neighborhood characteristics. Second, the Boston Foreclosure Report provides weekly listings of advertised sales and, since properties can be advertised in multiple months, duplicate listings for properties had to be removed from the database. We relied on borrower names and property addresses to identify and remove duplicate records. However, though we did manually check the remaining observations for possible duplicates, differences in the names or addresses from month to month (including errors and typos) may have prevented every duplicate record from being removed. Finally, a large proportion of the mortgagees listed in the foreclosure report could not be identified as HMDA reporters. In our efforts to keep our analysis of foreclosure trends analogous to that of origination trends, we excluded all foreclosures by non-hmda reporters from the analysis. Historical data on foreclosures were available only from 1995 to 1999. While it would be ideal to have foreclosure data back to 1994 in order to match the timeline of our loan originations, four years of data are adequate to conduct some trend analyses. Moreover, analyzing foreclosure data 15 The refinance market comprised between 35 and 44 percent of all HMDA-reported loan originations between 1994 and 1997, but jumped to 67 percent in 1998. September 2000 8

Analyzing Trends in Originations: A Case Study of the Boston Metro Area that are more recent than the origination data is reasonable, because trends in foreclosures lag trends in lending by at least one year, and generally by two or more years. There are several challenges to identifying subprime lenders in Boston Foreclosure Report. First, whereas both the HUD list of subprime lenders and the HMDA origination data contain lender identification numbers to facilitate matching of these lists, the BFR data does not contain this identifier. Thus, in order to determine whether each mortgagee in the BFR data was a subprime lender, we had to compare each mortgagee s name as it appeared in the property listing to a list of HMDA reporters and to HUD s list of subprime lenders. Aside from the shear volume of mortgagees listed (in excess of 7,000), this process was hampered by inconsistencies and errors in lender names. Second, BFR does not list the originating lender, only the current mortgagee servicing the mortgage. Since loan servicing is routinely acquired by other lenders after origination, the servicing mortgagee is not necessarily the originating mortgagee. Consequently, there is no way of knowing the exact role of the mortgagees listed in the BFR. It is possible that some of these lenders did not originate the loans on the foreclosed property causing loans to be misclassified as subprime or non-subprime. Since non-subprime lenders may service loans originated by subprime lenders (and, therefore, be listed as the mortgagee in BFR), it is possible that the levels and shares of subprime foreclosures are biased downward. 16 B. Findings 1. Overall Trends in Foreclosure While overall foreclosures in the Boston area declined by 30 percent between 1995 and 1999, subprime foreclosures increased by 154 percent. As a result, the share of foreclosures accounted for by subprime lenders rose from 3 to 11 percent, disproportionately high when compared to the subprime share of originations. The annual number of foreclosures decreased by about 30 percent in Boston between 1995 and 1999 (Exhibit 6). However, this overall decline masks very different trends in subprime and nonsubprime foreclosure rates. An overall decrease of 643 foreclosures of non-subprime lenders coincided with an increase in subprime lender foreclosures from 56 to 142 over the same period, an increase of 154 percent. As a result, the share of all foreclosures by subprime lenders increased from 3 percent to 11 percent during this period. As both the level and share of foreclosures attributable to subprime lenders have risen, their share has started to rise above their share of loan originations (Exhibit 7). The subprime lenders share of originations reached a high of 5 percent in 1997 and declined to 4 percent in 1998. But their share of foreclosures has continued to rise, reaching almost 7 percent in 1998 and 11 percent in 1999. Combined with the fact that it takes several years for newly originated loans to reach foreclosure, it looks as if foreclosures by subprime lenders are becoming disproportionately high compared to their overall share of lending activity. 16 For example, the NTIC study noted that court filings indicate that the Bank of New York commonly represents the interests of The Money Store in foreclosure proceedings in Cook County. September 2000 9

Analyzing Trends in Originations: A Case Study of the Boston Metro Area Exhibit 6 Lenders' of, 1995-1999 (Boston PMSA) Year Total Foreclosure by HMDA Reporters 1995 1,883 56 3.0% 1996 1,870 40 2.1% 1997 1,246 49 3.9% 1998 1,515 99 6.5% 1999 1,326 142 10.7% % Change -30% 154% 260% Source: Abt Associates Inc. tabulations of Boston Foreclosure Report data. Percentage of total originations and foreclosures by HMDA reporters only 12.0% 10.0% 8.0% 6.0% 4.0% 2.0% 0.0% Exhibit 7 HMDA Reporters' of Originations and (Boston PMSA) 1.7% 1.7% 3.0% 3.0% 2.1% 4.8% 3.9% 4.0% 6.5% 1994 1995 1996 1997 1998 1999 Originations 10.7% Source: Abt Associates analysis of data from HMDA and the Boston Foreclosure Report data. See Appendix tables 1 and 4. September 2000 10

Analyzing Trends in Originations: A Case Study of the Boston Metro Area 2. in Low-Income Areas The level and share of foreclosures accounted for by subprime lenders has grown rapidly in neighborhoods of all income levels. However, though the absolute growth in the number of subprime foreclosures was highest in low-income areas, the overall share of foreclosures by 1999 was roughly equal for all neighborhood types. Although the BFR data do not include the census tract in which the property is located, we were able to use the property addresses to create census tract information for 91% of the records, enabling us to link the foreclosure data with information on the income and racial composition of the neighborhoods for these records. The income categories for neighborhood income are based on criteria identical to those used for the origination data. 17 Exhibit 8 shows that the subprime lender foreclosures has been growing rapidly in neighborhoods of all income levels, with the highest growth in neighborhoods where median income is between 50 and 80 percent of the median income of the BMSA as a whole. But despite the somewhat unequal growth rates of subprime foreclosure by neighborhood income level, increases in the share of all foreclosures comprised by subprime lenders were roughly equivalent in the four income categories, with 1995 shares between 2 and 4 percent and 1999 shares between 10 and 12 percent. For each income category, the 1999 subprime share of foreclosures exceeded the 1999 subprime share of originations (Exhibit 9). 3. in minority neighborhoods While the volume and share of foreclosures by subprime lenders has grown rapidly in neighborhoods of all racial categories, neighborhoods with high proportions of minority residents now have much higher shares of foreclosures by subprime lenders. By 1999, among lenders reporting under HMDA, subprime lenders were accounting for about 15 percent of foreclosures in neighborhoods with high minority concentrations, higher than their share of originations in these areas. The volume of subprime foreclosures has increased by at least 150 percent in neighborhoods of all racial categories (Exhibit 10). In addition, the share of foreclosures attributable to subprime lenders more than tripled for all categories of racial composition between 1995 and 1999. However, the subprime share of foreclosures grew the most in communities with high proportions of non-white residents (Exhibit 11). The subprime share of foreclosures increased from 2 percent to 15 percent between 1995 and 1999 in neighborhoods with both high and very high concentrations of minorities. 17 That is, high income zip code areas have median incomes greater than 120 percent of the MSA median income, the moderate income category includes zip codes with median incomes between 80-120 percent of the MSA median income, the low-income neighborhoods have median incomes between 50 and 80 percent of the MSA median, and very low-income zip codes have median incomes below 50 percent of the area median. September 2000 11

Analyzing Trends in Originations: A Case Study of the Boston Metro Area 250% Exhibit 8 Growth in by Neighborhood Income Level 1995-1999 (Boston PMSA) 200% 150% 200% 158% 173% 100% 50% 50% 0% Very Low (<50%) Low (50-80%) Moderate (80-120%) Tract Median Household Income as Percent of Metro Area Median Source: Abt Associates Inc. analysis of Boston Foreclosure Report data. Note: See Appendix Table 4 for loan volumes. High (>120%) 14% 12% 10% 8% 6% 4% 2% Exhibit 9 of by Lenders by Neighborhood Income Level (Boston PMSA) 0% Very Low (<50%) Low (50-80%) Moderate (80-120%) Tract Median Household Income as Percent of Metro Area Median 1995 1996 1997 1998 1999 High (>120%) Source: Abt Associates Inc. analysis of HMDA and Boston Foreclosure Report data. Note: See Appendix2 for loan volumes September 2000 12

Analyzing Trends in Originations: A Case Study of the Boston Metro Area 250% Exhibit 10 Growth in by Neighborhood Racial Composition 1995-99 (Boston PMSA) 200% 150% 150% 175% 200% 161% 100% 50% 0% High (>50%) Moderately High (20-50%) Moderately Low (10-20%) Tract Median Household Income as Percent of Metro Area Median Source: Abt Associates Inc. analysis of Boston Foreclosure Report data. Note: See Appendix Table 4 for loan volumes. Low (<10%) Exhibit 11 of by Lenders by Neighborhood Racial Composition (Boston PMSA) 16% 14% 12% 10% 8% 6% 4% 2% 0% High (>50%) Moderately High (20-50%) Moderately Low (10-20%) Low (<10%) Tract Median Household Income as Percent of Metro Area Median 1995 1996 1997 1998 1999 Source: Abt Associates Inc. analysis of Boston Foreclosure Report data. Note: See Appendix Table 4 for loan volumes September 2000 13

Analyzing Trends in Originations: A Case Study of the Boston Metro Area 4. Age of Loan at Foreclosure The median age of foreclosed loans is only three years for subprime loans, compared to seven years for most other lenders, which is consistent with the greater risk of these loans at the time of origination. Another indication of the greater risk of subprime loans can be the age of the loan at foreclosure. Due to the high-risk nature of subprime loans, we would expect the time between loan origination and foreclosure to be shorter for subprime loans than for prime loans. By comparing the year of foreclosure to the year of origination, we can approximate the age of the loans listed in the Boston Foreclosure Report. 18 As expected, the median age of a foreclosed loan among the subprime lenders is lower than for other lenders (Exhibit 12). Over the entire period, the median age of foreclosed loans among subprime lenders was 3 years, compared to 7 years for non-subprime HMDA reporters. Overall, the age of non-subprime loans at foreclosures has declined over the five-year period studied from 7 years in 1995 to 5 years in 1999. This decline may be attributable to the strong refinancing boom of recent years that has greatly lowered average loan ages. Nonetheless, the age of subprime loans is still much lower than prime loans. The relatively short period from origination to foreclosure of these loans may be an indication that these loans are not affordable for borrowers even at origination. 18 Since BFR provides only the year and not the day or month of origination, the duration calculation is a somewhat crude approximation of the age of the loan. September 2000 14

Analyzing Trends in Originations: A Case Study of the Boston Metro Area Appendix 1. Activity in the Boston PMSA, 1994-1998 Table 1.a. Lenders' of HMDA Loans 1994 81841 1356 1.7% 1995 66272 1147 1.7% 1996 91424 2732 3.0% 1997 93684 4533 4.8% 1998 179486 7250 4.0% % Change 119.3% 434.7% Table 1.b. Lenders' of HMDA Loans, by Tract Median Income High-Income Medium-Income Low-Income Very Low-Income 1994 27,529 464 1.7% 44,028 688 1.6% 9,370 177 1.9% 792 16 2.0% 1995 21,415 279 1.3% 35,574 630 1.8% 8,388 197 2.3% 753 29 3.9% 1996 30,012 586 2.0% 48,951 1,614 3.3% 11,243 457 4.1% 1,056 59 5.6% 1997 30,832 1,018 3.3% 49,761 2,462 4.9% 11,750 847 7.2% 1,156 139 12.0% 1998 59,816 1,336 2.2% 96,691 4,045 4.2% 20,928 1,641 7.8% 1,879 188 10.0% % Change 117.3% 187.9% 119.6% 487.9% 123.4% 827.1% 137.2% 1075.0% Table 1.c. Lenders' of HMDA Loans, by Tract Racial Composition Less Than 10% Minority 10-20% Minority 20-50% Minority >50% Minority 1994 65,544 1,033 1.6% 10,094 178 1.8% 3,764 87 2.3% 2,437 58 2.4% 1995 52,003 840 1.6% 8,460 122 1.4% 3,401 67 2.0% 2,356 106 4.5% 1996 71,997 1,994 2.8% 11,681 371 3.2% 4,636 134 2.9% 2,983 217 7.3% 1997 73,814 3,210 4.3% 11,882 599 5.0% 4,736 284 6.0% 3,093 374 12.1% 1998 144,128 5,054 3.5% 22,035 1,000 4.5% 8,157 516 6.3% 5,065 641 12.7% % Change 119.90% 389.25% 118.30% 461.80% 116.71% 493.10% 107.84% 1005.17% Notes: a b activity is based on HMDA loan originations for the years 1994-1998. Composition by income reflects only the 511,924 records with information on median income. Racial composition reflects only the 512,266 records with information on racial composition. c The categories for income are defined as follows: Very Low: median tract income < 50% of the median income for the entire Boston PMSA Low: median tract income between 50% and 80% of the median income for the entire Boston PMSA Moderate: median tract income between 80% and 120% of the median income for the entire Boston PMSA High: median tract income > 120% of the median income for the entire Boston PMSA d For the racial composition breakdown, minority is defined as any non-white resident and includes whites of Hispanic origin September 2000 15

Analyzing Trends in Originations: A Case Study of the Boston Metro Area Appendix 2. Composition of HMDA-Reported Loans, by Loan Purpose Table 2.a. Level and of HMDA Loans, by Loan Purpose Loan Purpose Total Number of HMDA Loans Home Purchase Home Improvement Refinance Multifamily # % # % # % # % 1994 81,841 39,266 48.0% 6,893 8.4% 35,324 43.2% 358 0.4% 1995 66,272 35,892 54.2% 6,800 10.3% 23,271 35.1% 309 0.5% 1996 91,424 44,096 48.2% 6,950 7.6% 39,941 43.7% 437 0.5% 1997 93,684 46,074 49.2% 6,441 6.9% 40,682 43.4% 487 0.5% 1998 179,486 52,117 29.0% 6,165 3.4% 120,449 67.1% 755 0.4% September 2000 16

Analyzing Trends in Originations: A Case Study of the Boston Metro Area Appendix 3. Refinance Activity in the Boston PMSA, 1994-1998 Table 3.a. Lenders' of HMDA-Reported Refinance Loans Total Number of HMDA Refinance Loans Number of Refinance Loans 1994 35,324 825 2.3% 1995 23,271 640 2.8% 1996 39,941 1855 4.6% 1997 40,682 3209 7.9% 1998 120,449 5407 4.5% % Change 241.0% 555.4% Table 3.b. Lenders' of HMDA-Reported Refinance Loans, by Tract Median Income High-Income Medium-Income Low-Income Very Low-Income Number of HMDA Refinance Loans Number of Refinance Loans Number of HMDA Refinance Loans Number of Refinance Loans Number of HMDA Refinance Loans Number of Refinance Loans Number of HMDA Refinance Loans Number of Refinance Loans 1994 13239 264 2.0% 18491 450 2.4% 3309 98 3.0% 242 9 3.7% 1995 8510 146 1.7% 12243 364 3.0% 2297 112 4.9% 181 14 7.7% 1996 14205 389 2.7% 21289 1100 5.2% 4037 311 7.7% 338 41 12.1% 1997 14796 743 5.0% 21041 1736 8.3% 4346 593 13.6% 389 99 25.4% 1998 42862 1030 2.4% 64506 3042 4.7% 12076 1172 9.7% 885 137 15.5% % Change 223.8% 290.2% 248.9% 576.0% 264.9% 1095.9% 265.7% 1422.2% Table 3.c. Lenders' of HMDA-Reported Refinance Loans, by Tract Racial Composition Less Than 10% Minority 10-20% Minority 20-50% Minority >50% Minority 1994 29,233 641 2.2% 4,036 105 2.6% 1,348 42 3.1% 706 37 5.2% 1995 19,194 471 2.5% 2,598 71 2.7% 895 32 3.6% 569 62 10.9% 1996 32,639 1,369 4.2% 4,659 223 4.8% 1,603 75 4.7% 988 174 17.6% 1997 33,154 2,332 7.0% 4,648 378 8.1% 1,675 178 10.6% 1,106 283 25.6% 1998 99,144 3,823 3.9% 14,059 719 5.1% 4,464 349 7.8% 2,709 491 18.1% % Change 239.2% 496.4% 248.3% 584.8% 231.2% 731.0% 283.71% 1227.03% Notes: a activity is based on HMDA loan originations for the years 1994-1998. b Composition by income reflects only the with information on median income. Racial composition reflects only the records with information on racial composition. c The categories for income are defined as follows: Very Low: median tract income < 50% of the median income for the entire Boston PMSA Low: median tract income between 50% and 80% of the median income for the entire Boston PMSA Moderate: median tract income between 80% and 120% of the median income for the entire Boston PMSA High: median tract income > 120% of the median income for the entire Boston PMSA d For the racial composition breakdown, minority is defined as any non-white resident and includes whites of Hispanic origin September 2000 17

Analyzing Trends in Originations: A Case Study of the Boston Metro Area Appendix 4. Residential Foreclosure Activity in the Boston PMSA, 1995-1999 Table 4.a. Lenders' of Total By HMDA Reporters 1995 1,883 56 3.0% 1996 1,870 40 2.1% 1997 1,246 49 3.9% 1998 1,515 99 6.5% 1999 1,326 142 10.7% % Change -29.6% 153.6% Table 4.b. Lenders' of, by Tract Median Income High-Income Medium-Income Low-Income Very Low-Income 1995 393 15 3.8% 853 26 3.0% 384 9 2.3% 75 2 2.7% 1996 417 13 3.1% 857 19 2.2% 389 4 1.0% 50 2 4.0% 1997 322 14 4.3% 606 23 3.8% 198 7 3.5% 38 2 5.3% 1998 361 21 5.8% 753 52 6.9% 272 19 7.0% 32 0 0.0% 1999 362 41 11.3% 647 67 10.4% 219 27 12.3% 27 3 11.1% % Change -7.9% 173.3% -24.2% 157.7% -43.0% 200.0% -64.0% 50.0% Table 4.c. Lenders' of, by Tract Racial Composition Less Than 10% Minority 10-20% Minority 20-50% Minority >50% Minority 1995 1,235 41 3.3% 226 5 2.2% 161 4 2.5% 83 2 2.4% 1996 1,244 32 2.6% 233 3 1.3% 160 1 0.6% 76 2 2.6% 1997 877 35 4.0% 160 6 3.8% 83 1 1.2% 44 4 9.1% 1998 1,098 78 7.1% 186 10 5.4% 89 4 4.5% 45 0 0.0% 1999 1,002 107 10.7% 146 15 10.3% 73 11 15.1% 34 5 14.7% % Change -18.87% 160.98% -35.40% 200.00% -54.66% 175.00% -59.04% 150.00% a Notes: Compositions by income and race reflect only the 7255 records with census tract information. b The categories for income are defined as follows: Very Low: median tract income < 50% of the median income for the entire Boston PMSA Low: median tract income between 50% and 80% of the median income for the entire Boston PMSA Moderate: median tract income between 80% and 120% of the median income for the entire Boston PMSA High: median tract income > 120% of the median income for the entire Boston PMSA c For the racial composition breakdown, minority is defined as any non-white resident and includes whites of Hispanic origin September 2000 18