Federal Reserve

BULLETIN

A Profile of the 2010 Mortgage Market

HMDA covers all of the nation's leading home lenders as well as a large number of others. Banking institutions--commercial banks, savings institutions (savings and loans and savings banks), and credit unions--account for most of the reporting entities, although many mortgage companies are covered as well. For 2010, 7,923 institutions reported on their home-lending activity under HMDA: 3,818 commercial banks; 856 savings institutions; 2,041 credit unions; and 1,208 mortgage companies, 839 of which were not affiliated with a banking institution (table 1).

The number of reporting institutions has fluctuated over the years. Some of the fluctuation is due to changes in reporting requirements, including increases in the minimum asset level used to determine coverage.¹⁰ Mergers, acquisitions, and failures also account for some of the year-over-year changes. Finally, periodic changes in the number and geographic footprints of metropolitan areas influence reporting over time, as HMDA's coverage is limited to institutions that have at least one office in an MSA.

For 2010, the number of reporting institutions fell 2 percent from 2009, continuing a downward trend since 2006, when HMDA coverage included nearly 8,900 lenders. Among the types of reporters, the number of mortgage companies fell the most from 2009, decreasing 7 percent; since 2006, the number of mortgage companies has declined about 40 percent.

Reporting institutions vary greatly by both asset size and volume of reported mortgage lending activity. Most reporters are small, and many extend relatively few loans. For 2010, 53 percent of the banking institutions covered by HMDA had assets under $250 million, and 74 percent of them reported information on fewer than 100 loans; only 0.6 percent of these smaller banking institutions reported on more than 1,000 loans (table 2). Among all depository institutions, about 54 percent reported on fewer than 100 loans. Across different types of lenders, mortgage companies tend to originate larger numbers of loans on a per-reporter basis than the other institutions (38 percent of the mortgage companies reported more than 1,000 loans, a share more than five times that of depository institutions).

In any given year, institution failures and closings can affect the volume of reported loans and applications because some of the lenders that fail or were closed extended loans but did not report. Seventy-nine institutions that reported HMDA data for 2009 ceased operations and did not report lending activity for 2010.¹¹ Although it is not possible to know how many applications or loans these 79 institutions originated in 2010 before discontinuing operations, one can gauge their potential importance by measuring their lending activity in 2009. In the aggregate, these nonreporting companies accounted for only 0.2 percent of the 2009 loan or application records submitted under HMDA. Therefore, it seems highly unlikely that the 2010 data are affected in any meaningful way by the underreporting that may have occurred because these lenders did not report activity for the part of 2010 in which they may have made some loans.

Reporting institutions submitted information on 12.95 million applications for home loans of all types in 2010 (excluding requests for preapproval), down about 14 percent from 2009 and far below the 27.5 million applications processed in 2006, just before the housing market decline (data derived from table 3.A). The majority of loan applications are approved by lenders, and most of these approvals result in extensions of credit. In some cases, an application is approved, but the applicant decides not to take out the loan; for example, in 2010, about 5 percent of all applications were approved but not accepted by the applicant (data not shown in tables). Overall, about 61 percent of the applications submitted in 2010 resulted in an extension of credit (data derived from tables 3.A and 3.B), a share little changed from 2009.

The HMDA data also include information on loans purchased by reporting institutions during the reporting year, although the purchased loans may have been originated at any point in time. For 2010, lenders reported information on nearly 3.2 million loans that they had purchased from other institutions, a decline of nearly 25 percent from 2009. Finally, lenders reported on roughly 165,000 requests for preapproval of home-purchase loans that did not result in a loan origination (table 3.A); preapprovals that resulted in loans are included in the count of loan extensions cited earlier.

Lending for Home Purchase or Refinancing in 2010

Although relatively stable in the past two years, the volume of home-purchase lending has fallen sharply since 2006 (figure 1).¹² In June 2006, the peak month for home-purchase lending that year, nearly 712,000 home-purchase loans were extended, compared with only 326,000 such loans in June 2010, the most active month that year. On a yearly basis, the number of home-purchase loans reported by lenders covered by HMDA in 2010 was down nearly 9 percent from 2009 and was 62 percent lower than in 2006 (data derived from table 3.B).

Because of the extraordinary difficulties in the housing and mortgage markets, the federal government has taken several actions to support their recovery. One of the actions, the first-time homebuyer tax credit program, reduced the tax bill or increased the amount of refund for eligible homebuyers.¹³ The program was originally scheduled to end (or "sunset") on November 30, 2009, but was extended a few weeks to provide benefits to those eligible homebuyers who entered into binding contracts to purchase their homes by April 30, 2010, and closed the sales by June 30, 2010 (after the fact, the closing deadline was extended to September 30, 2010, but that extension affected only a small number of sales).

The first-time homebuyer tax credit program likely stimulated homebuying in 2009 as individuals sought to purchase their homes before the initial scheduled sunset date.¹⁴ The extension of the law until the end of June 2010 may help explain, in part, the increase in loan volume in the spring of 2010 and then the sharp falloff in the monthly flow of new home-purchase originations after that despite a decline in mortgage interest rates over the remainder of the year.¹⁵

To a greater degree than for home-purchase lending, the volume of refinance lending is aligned with changes in interest rates, expanding as mortgage rates fall and retrenching when rates rise. The interest rate environment in both 2009 and 2010 was generally quite favorable for well-qualified borrowers who sought to refinance, particularly in the second half of 2010, when the rate on 30-year fixed-rate mortgages fell to record lows. Nonetheless, compared with 2009, the number of reported refinancings was down about 14 percent (table 3.B). (Factors affecting the level of refinance activity in 2010 are explored in the later section "Factors Influencing Refinancing Activity in 2010.")

Back to section top

Non-Owner-Occupant Lending

Individuals buying homes either for investment purposes or as second or vacation homes are an important segment of the housing market in general, and in some areas of the country, they are particularly important. In the current period of high foreclosures and elevated levels of short sales, investor activity helps reduce the overhang of unsold and foreclosed properties. HMDA data help document the role of non-owner-occupant lending over time because the borrower's intended occupancy status is one of the reported items.¹⁶

As the boom in housing emerged in the first half of the past decade, the HMDA data showed a sharp increase in non-owner-occupant lending used to purchase one- to four-family homes (table 4). The volume of non-owner-occupant lending then fell sharply beginning in 2007 and remained at comparably low levels through 2010. In 2010, 76 percent fewer non-owner-occupant loans were extended than in 2005, the peak year for non-owner-occupant lending. The number of non-owner-occupant loans in 2010 was little changed from that in 2009.

As shown in table 4, the post-2007 decline in non-owner-occupant lending has been more severe than that in owner-occupant lending. Between 2000 and 2005, the share of non-owner-occupant lending used to purchase one- to four-family homes rose, increasing over this period from about 9 percent to 16 percent (data derived from table 4).¹⁷ Since 2005, the share has fallen, dropping to about 11 percent in both 2009 and 2010. Although diminished since the middle of this decade, in both the volume of lending and as a share of lending, non-owner-occupant lending continues to be an important aspect of the mortgage market.

As noted, the relative importance of non-owner-occupant lending varies from place to place. In some places, such as resort towns, non-owner-occupant lending reflects the activity of both investors and purchasers of second homes. In other areas, most of the non-owner-occupant homebuying is by investors seeking to buy units for year-round rental. The HMDA data provide an opportunity to explore the geographic variation in non-owner-occupant lending across geographies, although it is not possible to distinguish between investors and second-home buyers.

For the analysis here, we calculated the non-owner-occupant share of home-purchase lending in each county nationwide in 2010 (figure 2). Many of the counties with elevated non-owner-occupant lending rates are resort locations such as portions of northern New England, Michigan, Colorado, and some coastal areas.¹⁸ Other areas may have elevated non-owner-occupant activity due to investors purchasing homes in markets that have experienced significant declines in home values, such as the "sand states" of Arizona, California, Florida, and Nevada. The sharply reduced values of properties involved in short sales or foreclosures have afforded investors and others opportunities to purchase non-owner-occupied homes in these areas.

Nevertheless, the decline in the volume of non-owner-occupant lending that has been observed nationally has affected almost all geographic areas. In all but a handful of MSAs, the percentage decline in non-owner-occupant lending between 2005 and 2010 exceeded the decline in owner-occupant lending (figure 3). Non-owner-occupant lending has fallen the most in the MSAs that experienced the largest declines in owner-occupant lending.

Further analysis suggests that at least some of the decline in non-owner-occupant lending stems from the locations where such loans were concentrated rather than an overall change in the national market for such loans. We selected an analysis group of census tracts in the top 5 percent of the incidence of total one- to four-family non-owner-occupant lending in MSAs in the sand states (where non-owner-occupant lending has been cited as a particular problem) in 2005. We compared the decline in lending in these tracts between 2005 and 2010 with the lending changes in all other tracts in the sand state MSAs.

Overall, lending in the analysis group declined 74 percent between 2005 and 2010, although non-owner-occupant lending fell more in these tracts (78 percent) than owner-occupant lending (71 percent). In contrast, overall lending fell significantly less in tracts where non-owner-occupant lending had not been concentrated (59 percent), with non-owner-occupant lending again experiencing a relatively larger decline (70 percent) than owner-occupant lending (58 percent). It cannot be determined from these results whether characteristics of the tract neighborhoods or the high presence of non-owner-occupant lending led to the excessive decline.

Types of Loans

As noted, the total number of loans to purchase homes has fallen sharply since the height of the housing boom in 2005 and 2006, when lenders extended about 7 million loans in each of those years (table 3.B). Although the total number of home-purchase loans has fallen substantially since then, virtually all of the decline has involved conventional lending; the volume of nonconventional home-purchase loans (sometimes referred to as "government backed" loans)--including loans backed by insurance from the FHA or by guarantees from the VA, the Farm Service Agency (FSA), or the Rural Housing Service (RHS)--has increased markedly since the mid-2000s. From 2006 to 2009, the total number of reported conventional home-purchase loans fell 77 percent, while the number of nonconventional home-purchase loans more than tripled (table 4). Although the number of nonconventional home-purchase loans fell some from 2009 to 2010, such loans still accounted for 46 percent of the home-purchase loan market in 2010, down marginally from a 48 percent share in 2009 but still much greater than the 8 percent share in 2006.

Nonconventional loans are a major component of the overall home-purchase loan market, but they play a much smaller role in certain segments of the market. For example, nonconventional loans accounted for only about 1 percent of the loans extended to non-owner occupants for the purchase of a home in 2010. Also, nonconventional loans made up a relatively small share (about 25 percent) of the loans used to purchase manufactured homes (table 5).

As in the home-purchase loan market, nonconventional lending has also garnered a larger share of the refinance market in the past few years, although the number of conventional loans used for refinancing still exceeds that of nonconventional loans by a wide margin (table 4). In 2006, conventional loans used for refinancing outnumbered nonconventional loans 48 to 1; in 2010, the proportion was about 6 to 1.

The increase in nonconventional home-purchase and refinance lending reflects several factors, such as increased loan-size limits allowed under the FHA and VA lending programs and reduced access (including more-stringent underwriting and higher prices) to conventional loans, particularly those that allow the borrower to finance more than 80 percent of the property value. (These factors and their role in 2010 lending are discussed in more detail in a later section, "The Continuing Role of Government in the Mortgage Market.")

Back to section top

The Private Mortgage Insurance Market

Historically, mortgage lenders have required a borrower to make a down payment before they would extend a loan to buy a home or refinance an existing mortgage. In the conventional loan market, lenders typically have required that a borrower make a down payment of at least 20 percent of a home's value unless the borrower received some type of third-party backing, such as mortgage insurance.

Private mortgage insurance (PMI) emerged in the 1950s alongside the longstanding FHA and VA loan programs to help bridge the gap between lenders reluctant to extend mortgages with high loan-to-value (LTV) ratios and consumers interested in borrowing more than 80 percent of the underlying home's value. For a borrower seeking a conventional loan with a low down payment, the lender can require that the borrower purchase mortgage insurance from PMI companies to protect the lender against default-related losses up to a contractually established percentage of the principal amount.

Over the years, PMI-backed loans became a significant part of the mortgage market. As a form of protection for lenders against losses from defaulting borrowers, PMI competes with FHA insurance and VA loan guarantees. Thus, the relative attractiveness of PMI at any point in time is closely related to FHA and VA underwriting and pricing decisions and the sizes of the loans these government agencies may back. PMI also competes against the willingness of lenders to bear the risk of loss through self-insurance by extending a first-lien mortgage with little or no down payment in conjunction with a junior-lien mortgage (often referred to as a "piggyback" loan). Historically, the annual volume of PMI issuance has varied in response to these competitive pressures and to the overall level of mortgage activity in any given year.

In 1993, the Mortgage Insurance Companies of America asked the FFIEC to process data from the largest PMI companies on applications for mortgage insurance and to produce disclosure statements for the public based on the data and timed to be released with the HMDA data.¹⁹ The PMI data largely mirror the types of information submitted by lenders covered by HMDA. However, because the PMI companies do not receive all of the information about a prospective loan from the lenders seeking insurance coverage, some items reported under HMDA are not included in the PMI data. In particular, loan pricing information and requests for preapproval are unavailable in the PMI data.

The seven companies that reported data for 2010 dominate the PMI industry.²⁰ Thus, these data cover the vast majority of PMI written in the United States, allowing for meaningful analysis of these data alongside the HMDA data.²¹ For 2010, the seven PMI companies reported on nearly 370,000 applications for insurance leading to the issuance of 260,000 insurance policies, down from about 636,000 applications and 367,000 policies in 2009 (table 6). Both the 2009 and 2010 volumes were substantially smaller than the totals reached in 2002 and 2003, when PMI issuance was about 2 million policies a year. Overall, 61 percent of the PMI policies issued in 2010 covered home-purchase loans, and the remainder covered refinance mortgages (home-improvement loans are classified as refinance loans by the PMI reporters). Virtually all of the applications for PMI policies issued involved site-built properties; less than 0.04 percent of the policies involved manufactured homes. About 10 percent of PMI insurance applications were denied in 2010, down from about 12 percent in 2009 but still substantially higher than in 2006 and 2007, when only about 2 percent of the requests for insurance were turned down (data not shown in tables).²²

The large reduction in PMI activity reflects several factors, including reduced demand stemming from a sharp fall in homebuying activity and higher prices relative to alternatives, as well as tighter underwriting adopted by the PMI companies in response to elevated claims and losses experienced during the recent recession and the ongoing recovery.²³ The roles of these various factors can be seen from the memo items in the last seven columns of table 6, which focus on owner-occupant site-built lending. Taken in isolation, PMI rose as a share of conventional lending from 2006 to 2007 and then fell back sharply in 2009 and further still in 2010. Some of this change reflects variation in the share of borrowers with high-LTV loans. However, as can be seen from the table, much of the change, particularly since 2008, reflects substitution among high-LTV credit enhancement alternatives, including nonconventional FHA and VA loans and junior-lien piggyback loans. Indeed, since 2008, the share of total home-purchase loans covered by one of these enhancements has remained quite stable (last column of table 6). Thus, the record low number of PMI policies issued in 2010 likely paints a very misleading picture regarding high-LTV lending. The steadily rising share of the loan market covered by some sort of credit enhancement evident in the last column of table 6 suggests that high-LTV loans, perhaps driven by a rising portion of the market that is composed of first-time homebuyers, may be at record high rather than record low levels.

Back to section top

Junior-Lien Lending

Junior-lien loans can be taken out either in conjunction with the primary mortgage (a piggyback loan) or independently of the first-lien loan. As noted in the previous section, piggyback loans can be used by borrowers to avoid having to pay for private or government mortgage insurance. Similarly, piggyback loans can also be used to reduce the size of the first-lien loan to be within the size limits required by Freddie Mac or Fannie Mae without requiring a larger down payment by the borrower. Junior-lien loans that are taken out independently can be used for any number of purposes, including to finance home-improvement projects or, in the case of open-ended home equity lines of credit (HELOCs), to provide a readily available and relatively cheap source of credit. Under the regulations that govern HMDA reporting, most of these stand-alone junior-lien loans are not reported.²⁴ Most piggyback loans, however, must be reported. Thus, the volume of junior-lien lending reported under HMDA may be more indicative of the volume of piggyback lending than of junior-lien lending as a whole.

Before the financial crisis and the collapse in home values, when the use of piggyback loans was more common and the size limits on Fannie Mae and Freddie Mac purchases were lower, many more junior-lien loans were reported in the HMDA data. For example, in 2006, which represented the high-water mark for junior-lien lending, over 1.3 million conventional junior liens used for the purchase of owner-occupied properties were reported under HMDA, and another 1 million conventional junior-lien loans were taken out to refinance loans backed by owner-occupied properties (table 7). Virtually all such lending was conventional; fewer than 1,000 loans involved government-backed programs. As the elevated credit risk associated with high-LTV-ratio lending became apparent during the Great Recession and its aftermath, underwriting tightened and junior-lien loans became difficult to obtain or were no longer made available. The number of junior-lien loans for the purchase of owner-occupied homes reported under HMDA fell by more than one-half in 2007, dropped sharply again in each of the ensuing years, and decreased somewhat to about 40,000 such loans in 2010. The number of junior-lien loans used for refinancing also fell substantially starting in 2007 and continued to fall, reaching almost 88,000 in 2010. Substantial declines were also observed in the number of junior-lien loans backed by non-owner-occupied properties, whether the loans were for home purchase or refinancing.

The category in which the number of junior-lien loans reported in the HMDA data has declined the least has been junior-lien loans for home-improvement purposes. In 2010, almost 80,000 junior-lien loans were used for home improvement. While this number was down 11 percent from 2009 and 86 percent from 2005, the decline was less steep than that observed for other types of junior-lien lending. As a result, junior-lien loans used for home improvement accounted for 37 percent of junior-lien loans reported under HMDA.

Back to section top

Loan Sales

Among the information included in the annual HMDA data is the type of purchaser for loans that are originated and sold during the year. For purposes of reporting, lenders are provided with nine types of purchasers that may be used to classify loan sale activity. Broadly, these purchaser types can be broken into those that are government related--Ginnie Mae, Fannie Mae, Freddie Mac, and Farmer Mac--and those that are not.²⁵ Ginnie Mae and Farmer Mac focus on loans backed directly by government guarantees or insurance, primarily FHA-insured or VA-guaranteed loans. The government-sponsored enterprises (GSEs) Fannie Mae and Freddie Mac are focused on conventional loans that meet the underwriting standards established by those entities.

Overall, 80 percent of the first-lien home-purchase and refinance loans for one- to four-family properties originated in 2010 were reported as sold during the year (data not shown in tables). The share of originations that are sold varies some from year to year and by type and purpose of the loan (table 8).²⁶ For example, about 70 percent of the conventional loans extended in 2010 for the purchase of owner-occupied one- to four-family dwellings were sold that year. In contrast, about 93 percent of the nonconventional loans used to purchase owner-occupied homes were reported as sold in 2010. The share of conventional loans made to non-owner occupants that are reported as sold is notably smaller than that of such loans made to owner occupants, as is the share of loans extended for the purchase of manufactured homes.

Although one of the few sources of information on loan sales, the HMDA data tend to understate the importance of the secondary market. HMDA reporters are instructed to record loans sold in a calendar year different from the year originated as being held in portfolio, leading the reported loan sales to understate the proportion of each year's originations that are eventually sold. (We deal with this issue in more detail in the later section "The Continuing Role of Government in the Mortgage Market.")

Back to section top

Borrower Incomes and Loan Sizes

Under the provisions of HMDA, lenders report the loan amount applied for and the applicant income that the lender relied on in making the credit decision, if income was considered in the underwriting decision. The vast majority of loan applications and loans reported under HMDA include income information. For example, in 2010, income information was not reported for only about 1 percent of the borrowers purchasing a home with a nonconventional loan and for about 3 percent of those using a conventional loan (data not shown in tables). Income information is not reported more often for refinance loans, particularly those that are nonconventional (about one-third of the FHA loans and two-thirds of the VA loans), likely because of streamlined refinance programs that do not require current income to be considered in underwriting.

While the available information on amounts borrowed and income can be evaluated in many ways, here the focus is on patterns by loan product. For home-purchase or refinance lending, borrowers using FHA and VA loans have lower mean or median incomes than other loans despite the fact that the FHA (and VA) loan limits were increased substantially in 2008, allowing the program to be used much more widely than by the lower- and moderate-income households that have been the traditional focus of the program (table 9). For example, in 2007, the year before the increase in loan limits, about 7 percent of FHA borrowers had incomes of $100,000 or more, while in 2010, the share increased to 15 percent. Overall, in 2010, the median incomes for FHA, VA, and conventional loan borrowers were $55,000, $68,000, and $110,000, respectively (data for only 2010 shown in tables).

Loan amounts also differ across loan types, with FHA or VA loans, on average, being smaller than "other" loans (table 10). However, an upward shift in the distribution of loan amounts for both FHA and VA home-purchase loans occurred in the past couple of years, continuing into 2010 (data for only 2010 shown in tables). The shift likely reflects the same forces that are changing the distribution of borrower incomes.

Back to section top

Application Disposition, Loan Pricing, and Status under the Home Ownership and Equity Protection Act

For purposes of analysis, loan applications, loans, and requests for preapproval reported under HMDA can be grouped in many ways. Every loan application and request for preapproval reported in 2010 can be categorized into 25 distinct product categories characterized by type of loan and property, purpose of the loan, and lien and owner-occupancy status (tables 11 and 12). Each product category contains information on the number of total and preapproval applications, application denials, originated loans, loans with prices above the reporting thresholds established by HMDA reporting rules for identifying higher-priced loans, loans covered by the Home Ownership and Equity Protection Act of 1994 (HOEPA), and the mean and median annual percentage rate (APR) spreads for loans reported as higher priced.

Disposition of Applications

As noted, the 2010 HMDA data include information on nearly 13 million loan applications, about 85 percent of which were acted upon by the lender (data derived from table 11). Patterns of denial rates are largely consistent with what has been observed in earlier years.²⁷ Denial rates on applications for home-purchase loans are notably lower than those observed on applications for refinance or home-improvement loans. Denial rates on applications backed by manufactured housing are much higher than those on applications backed by site-built homes. For example, the denial rate for first-lien conventional home-purchase loan applications for owner-occupied site-built properties was 15.1 percent in 2010, compared with a denial rate of 55 percent for such applications for owner-occupied manufactured homes.

In addition to the application data provided under HMDA, nearly 443,000 requests for preapproval were reported as acted on by the lender in 2010 (table 12). About 26 percent of these requests for preapproval were denied by the lender. Not surprisingly, the number of requests for preapproval is down substantially from the levels recorded at the height of the housing boom. In 2006, covered institutions reported that they received nearly 1.2 million requests for preapproval upon which they took action (data not shown in tables).

Back to section top

Rule Changes Related to Higher-Priced Lending

The rules governing whether a loan is classified as higher priced under HMDA were changed in 2008, with implementation affecting loan classifications for applications after October 1, 2009. All loans reported in the 2010 HMDA data, regardless of the date of application, are covered under the new rules. The purpose of the rule change was to address concerns that had arisen about the distortive effects of changes in the interest rate environment on the reporting of higher-priced lending under the original methodology.²⁸ Under the original methodology, changes in underlying market rates of interest, particularly a steepening or flattening of the yield curve, could result in two loans of equivalent credit and prepayment risk being classified differently under HMDA as higher priced or not at different points in time, an outcome that was unintended.²⁹

To address the distortions arising from the method used to classify loans as higher priced or not, the price-reporting rules under HMDA were modified. Lenders are now required to compare the APR on the loan with the "average prime offer rate" (APOR) for loans of a similar type (for example, a 30-year fixed-rate mortgage). The APOR, which is published weekly by the FFIEC, is an estimate of the APR on loans being offered to high-quality prime borrowers based on the contract interest rates and discount points reported by Freddie Mac in its Primary Mortgage Market Survey (PMMS).³⁰ If the difference is more than 1.5 percentage points for a first-lien loan or more than 3.5 percentage points for a junior-lien loan, then the loan is classified as higher priced and the rate spread is reported. Since APORs move with changes in market rates and are product specific, it is anticipated that the distortions that existed under the old methodology will be overcome.

Back to section top

The Incidence of Higher-Priced Lending

The data show that the incidence of higher-priced lending across all products in 2010 was about 3.2 percent (table 11).³¹ The incidence varies across loan types, products, and purposes. First, in almost all cases, nonconventional loans have a lower incidence of higher-priced lending than do comparable conventional loan products, although the differences in incidence are much smaller than in the period when many conventional loans were subprime or near prime. In 2010, among first-lien home-purchase loans for site-built homes, 3.3 percent of conventional loans had APRs above the price-reporting threshold, versus 1.3 percent of nonconventional loans. Second, with few exceptions, first-lien loans have a lower incidence of higher-priced lending than do junior-lien loans for the same purposes. For example, in 2010, the incidence of higher-priced lending for conventional first-lien refinance loans was 1.3 percent, whereas for comparable junior-lien loans it was 11.9 percent. This relationship is found despite the fact that the threshold for reporting a junior-lien loan as higher priced is 2 percentage points higher than it is for so reporting a first-lien loan. Third, manufactured-home loans exhibit the greatest incidence of higher-priced lending across all loan categories, a result consistent with the elevated credit risk associated with such lending. For 2010, 80 percent of the conventional first-lien loans used to purchase manufactured homes were higher priced.

Back to section top

Rate Spreads for Higher-Priced Loans

Although there is considerable variation across loan products in the incidence of higher-priced lending, the variation across products in mean and median APOR spreads as reported in the HMDA data is much smaller. For example, for 2010, the mean APOR spread reported for higher-priced conventional first-lien loans for the purchase of an owner-occupied site-built home was about 2.5 percentage points, compared with about 1.8 percentage points for higher-priced first-lien nonconventional loans used for the same purpose (table 11).

It is worth noting that the vast majority of nonconventional loans reported as higher priced in 2010 exceeded the HMDA price-reporting thresholds by only a small amount: Specifically, 80 percent of the higher-priced nonconventional first-lien home-purchase loans had reported spreads within 50 basis points of the threshold. By comparison, only about 40 percent of the comparable conventional loans reported as higher priced had prices this close to the margin of reporting. In contrast, the share of higher-priced nonconventional refinancing loans with APORs close to the margin of reporting (39 percent) is a little less than the share of higher-priced conventional refinancing loans with such APORs (about 42 percent).

As expected, consistent with the higher reporting threshold of junior-lien lending, higher-priced junior-lien loan products have higher mean and median APOR spreads than do higher-priced first-lien loans. Higher-priced loans for manufactured homes differ from other loan products in that they generally have the highest mean spreads. In 2010, the typical higher-priced conventional first-lien loan to purchase a manufactured home had a reported spread of about 5.2 percentage points, compared with an average spread of roughly 2.5 percentage points for comparable higher-priced loans for site-built properties.

Back to section top

HOEPA Loans

The HMDA data indicate which loans are covered by the protections afforded by HOEPA. Under HOEPA, certain types of mortgage loans that have interest rates or fees above specified levels require additional disclosures to consumers and are subject to various restrictions on loan terms.³² For 2010, 655 lenders reported extending nearly 3,400 loans covered by HOEPA (table 11; data regarding lenders not shown in tables). In comparison, 1,153 lenders reported on about 6,500 loans covered by HOEPA in 2009. In the aggregate, HOEPA-related lending made up less than 0.1 percent of all of the originations of home-secured refinancings and home-improvement loans reported for 2010 (data derived from tables).³³

Back to section top

Factors Influencing Refinancing Activity in 2010

As discussed earlier, the APOR for a 30-year fixed-rate mortgage fell sharply at the end of 2008 and into 2009, and then it fell to well under 5 percent in 2010 (figure 1). Moreover, these rate declines appear to have sparked elevated refinance activity in early 2009 and late 2010. Still, overall refinance activity in both 2009 and 2010 appears low compared with what might have been expected given the sharp decline in interest rates. For example, interest rates last fell sharply in the early 2000s, and refinance volume peaked in 2003 at over 15 million loans, more than the combined refinance volume in 2009 and 2010 (table 3.B).

One explanation for subdued refinance activity is that lenders may be less willing or less able to take risk than earlier in the decade. The Federal Reserve's quarterly Senior Loan Officer Opinion Survey on Bank Lending Practices indicates that lenders have tightened credit standards during the past few years.³⁴ Lenders could also be adjusting prices in light of perceptions that borrowers with the same nominal characteristics (credit score, for example) pose more risk now than they did several years ago because of the substantially weaker and more uncertain conditions in employment and housing markets. Lenders may also be pricing risk more stringently because they are passing on certain fees from the GSEs. In 2008, both Fannie Mae and Freddie Mac started charging additional fees ("loan-level pricing adjustments") on loans to borrowers in credit score and LTV ranges in which they had not charged such fees before. In addition to these factors, the increased fees and tighter underwriting by PMI companies noted earlier and the increased presence of junior-lien loans--which must be closed, refinanced, or resubordinated--relative to the past may also be impeding refinance activity.

Subdued refinance activity may also stem from the fact that the financial standing of many borrowers has been undermined by sharp house price declines and the associated loss of home equity, especially for those residing in Arizona, California, Florida, Michigan, and Nevada, where home prices fell more than 20 percent from the end of 2006 to the end of 2009, according to the Federal Housing Finance Agency (FHFA) home price index.³⁵ A borrower with little or no home equity may need to pay down his or her loan balance substantially before being able to qualify for a refinance, which could be difficult.

That said, a few programs have been introduced in recent years to help facilitate refinancing for those with little or no equity. Perhaps most notable is the Home Affordable Refinance Program. To be eligible, borrowers must be current on their payments, and their loans must have been originated before June 2009, be backed by the GSEs, and have balances that do not exceed 125 percent of the respective home values. Thus far, the program has had less of an effect than initially expected, perhaps because of some issues raised previously, such as getting junior-lien holders to agree to resubordinate their loans under the new refinance loan.³⁶

To help describe refinance activity in more detail than is possible with the HMDA data, we draw on a relatively new data source--the FRBNY Consumer Credit Panel/Equifax. The panel is a nationally representative longitudinal database of individuals with detailed information, at a quarterly frequency beginning in 1999, on consumer and mortgage debt and loan performance drawn from the credit records collected and maintained by Equifax, one of the three national credit bureaus.³⁷ The data include three key pieces of information with respect to this analysis: (1) details on each mortgage outstanding for a given consumer, including the year of origination and current balance; (2) each consumer's credit score as of the end of 2009; and (3) each consumer's geographic location at the level of the census block (a subunit of a census tract).³⁸

Refinance mortgage loans are not explicitly identified in the credit bureau data, but because we can follow a given mortgage borrower over time, we can infer whether that borrower refinanced his or her mortgage during any particular period. Estimates of "refinance rates" (the shares of borrowers that refinanced their mortgages) during 2010 are provided by credit score, geography, and year of loan origination for those with mortgages outstanding as of the end of 2009 (table 13, top panel). To simplify the analysis, we focus on consumers who had exactly one closed-end mortgage with an outstanding balance of at least $50,000 as of the end of 2009 and stayed in the same census block over the course of 2010 (a proxy for not having changed residence; we simply want to omit from the analysis those who moved).³⁹ We then look at their mortgage accounts at the end of 2010 and classify consumers as having refinanced during 2010 if they (1) opened at least one closed-end mortgage in 2010 and (2) no longer had a positive-balance mortgage with an origination date matching that of the mortgage that was outstanding at the end of 2009.

Estimated refinance rates in 2010 were highest among consumers with pristine credit scores (820 or higher) whose loans were originated between 2006 and 2008--years with relatively high interest rates.⁴⁰ Within these origination years, lower credit scores were associated with much lower refinance rates within both groups of states. For example, in "other" states--those states that did not experience the largest declines in home values--refinance rates for consumers with credit scores of 680 to 719 were less than half of those for consumers with the highest credit scores.

Estimated refinance rates are also generally lower for borrowers in the five states that experienced house price declines of 20 percent or more ("sharpest declines") within each score group, especially for loans originated between 2005 and 2007--the time that house prices peaked. Those who purchased homes without significant down payments or reduced their equity substantially through refinancing during this period would have been most affected by declining house prices. Also, the difference in refinance rates tends to rise as credit scores fall, suggesting that low equity seems to compound the problem of lower scores in efforts to refinance. For example, among those with loans originated in 2006, the estimated refinance rate for consumers with pristine credit scores is nearly 25 percent in the other states, compared with roughly 17 percent in the states with the sharpest declines, while for the group with the second-highest score (720-819), this difference becomes more pronounced--about 19 percent versus 9 percent.

In contrast, analogous estimates for 2003 in the bottom panel of table 13 indicate that refinance rates that year did not vary systematically across state groups and did not fall with credit scores until scores dropped below 680. In other words, the current pattern of refinance activity across state and credit score groups does not appear to be explained by historical precedent. The findings overall provide evidence consistent with the view that losses of home equity, weak economic conditions, and tighter underwriting (whether due to decreased appetite for risk or an increased assessment of risk) damped refinance activity in 2010.⁴¹

We can use the estimates in table 13 to make an inference about the share of mortgagors at the end of the 2009 who would have refinanced during 2010 if home equity was not an impediment and underwriting was similar to what it was in 2003 (that is, similar across credit score groups except for the group with scores below 680). Specifically, if we assume that refinance rates for each state-score-year group were identical to the refinance rates for those with pristine scores in other states and that, as in 2003, refinance rates for those with scores below 680 are about 60 percent of what they were for pristine score types, then the overall refinance rate in 2010 would have been just over 16 percent instead of just under 11 percent.⁴²

This "counterfactual" refinance rate might be conservative since it abstracts away from the problems posed by the increased incidence of junior liens, noted earlier. However, we also estimated refinance rates for those with a positive-balance HELOC and found nearly identical refinance rates as those shown in table 13.⁴³ The counterfactual rate might also be conservative because declining house prices affected borrowers in the other states at least to some extent, but, notably, we found refinance rates for those in the subset of states where prices have not declined since 2006 to be very similar to those for the other states as a whole (data not shown in tables). Overall, this exercise suggests that refinance rates could have been just over 5 percentage points, or just over 50 percent, higher in the absence of home equity problems and underwriting changes. Applying that number to the HMDA data implies that roughly an additional 2.3 million first-lien owner-occupant refinance loans would have been made during 2010 on top of the roughly 4.5 million such loans that were in fact originated (data derived from table 7).

Back to section top

The Continuing Role of Government in the Mortgage Market

The HMDA data for 2008 and 2009 showed that the share of new mortgage loans either explicitly or implicitly guaranteed by the federal government rose dramatically from 2006. This increased government role continued in 2010, with the share of loans that were originated through the FHA, VA, and, to a much lesser extent, FSA or RHS programs, or that were owned outright or in mortgage pools guaranteed by Fannie Mae or Freddie Mac, remaining about the same as it was in 2009. This section discusses the underlying causes of this trend. To facilitate our analysis, we employ a revised data set designed to correct for one of the limitations in the HMDA reporting system.

Under HMDA reporting rules, all loans originated under the FHA, VA, FSA, or RHS programs must be identified as such.⁴⁴ However, loans placed in pools that are guaranteed by or sold to the GSEs Fannie Mae and Freddie Mac are identified only if they are sold directly to the GSEs or directly placed in a pool during the same year of the loan origination. The HMDA data therefore tend to undercount loans ultimately sold to the GSEs for two reasons. First, sales can take place in a year subsequent to origination, especially among loans originated during the fourth quarter. Second, lenders may not sell loans directly to the GSEs but instead may sell them to other financial institutions that form mortgage pools for which investors subsequently obtain GSE credit guarantees.

For the analysis in this section, we adjust the HMDA data to attempt to correct for the undercount of GSE loans. First, in addition to reporting their mortgage originations, financial institutions are also required to report their loan purchases and the types of purchasers if the loans are resold. Using information on loan size, location, date of origination, and date of purchase, we were able to match more than 50 percent of the loans that were originated from 2006 to 2010 and then sold to other financial institutions to the records for the same loans in the loan purchase files. From those matched loans, we were then able to obtain the ultimate loan dispositions from the filings of loan purchases. Of the portion we were unable to match, most were originated (and purchased) by one large organization, which supplied us with the aggregate disposition of the purchased loans. For those sold loans that we were still unable to match, we assumed that the distribution of the ultimate disposition matched that of similar loans that we could match.

Second, to address the undercount of GSE loans originated in October through December of each year, we used an imputation formula based on the allocation of loans originated in the preceding September and the following January to assign the ultimate disposition of conventional loans.⁴⁵ The imputation was conducted separately for the largest mortgage originators and took account of the characteristics of the loan, including size and location.

Figure 4 illustrates the changing structure of the mortgage market between 2006 and 2010 using our adjusted data for the four major loan types reported under HMDA. It groups first-lien site-built mortgages into four distinct categories: (1) loans insured by the FHA, backed by the VA, or issued or guaranteed by the FSA or RHS ("nonconventional"); (2) conventional loans sold to Fannie Mae or Freddie Mac or placed in pools guaranteed by them ("GSE"); (3) conventional loans sold to an affiliate or held in the portfolio of the originating lender ("portfolio"); and (4) all other conventional loans, including those sold into the private securitization market or to unaffiliated institutions ("other"). Panels 4.A, 4.B, and 4.C show patterns for owner-occupant home-purchase, refinance, and home-improvement loans; panel 4.D shows patterns for all non-owner-occupant loans regardless of purpose.⁴⁶

Our adjusted data show a greater role for the GSEs than that implied by the raw HMDA data. The raw HMDA data reported by lenders show that 39.6 percent of owner-occupant refinance loans originated in 2010 were reported as sold directly to the GSEs or placed in a mortgage pool guaranteed by them (data derived from tables 7 and 8); our revised data imply that the "correct" figure is likely to be much higher (over 55 percent).

The data in figure 4 show that the subprime-based private securitization market declined at the end of 2006 and throughout 2007, while the GSEs gained market share. Portfolio and nonconventional market shares remained relatively constant until the end of 2007. The years 2008 through 2010 show a different dynamic, with nonconventional home-purchase market share rising dramatically and then remaining constant before dropping somewhat in 2010. The GSEs play a much more prominent role in the refinance market; their share rose dramatically at the beginning of 2008, fell through August, and then rose again into 2009 and 2010. Portfolio and other lending dropped precipitously from 2007 to 2009 before increasing somewhat in 2010, particularly in the refinance market.

These changing patterns reflect the actions of a number of players. Nonconventional lending has traditionally focused on the high-LTV market, offering investors mortgage insurance protection against borrower default. PMI companies also offer similar insurance for high-LTV conventional loans, with PMI (or some other credit enhancement) required by statute for loans with LTVs above 80 percent that are sold to the GSEs. Lenders can also choose to forgo PMI and (1) hold the loan directly or (2) issue a junior-lien piggyback loan for the portion of the loan above 80 percent and still sell the 80 percent loan to the GSEs. The choice among PMI, public mortgage insurance, or a piggyback loan is likely to be made by borrowers (and lenders) based on the relative pricing and underwriting standards of the PMI and the nonconventional loan products. Prices and underwriting established by purchasers in the secondary market also matter. Both GSEs charge fees for loans they purchase or guarantee, with the fees varying by LTV and credit quality and subject to change over time. The GSE, FHA, and VA programs are also subject to statutory limits on loan size, which can and have been changed. Finally, the willingness of financial institutions to hold mortgages in portfolio is likely to be sensitive to their cost of funds, their capital position, and other factors.

Many of these items have changed over the past five years and likely influenced lending outcomes, as described in previous articles. Relative to previous years, there was relatively little change in 2010. The most notable event, discussed earlier, was the expiration of the first-time homebuyer tax credit program. The program--which, in an earlier article, we estimated that one-half of the home-purchase loans in 2009 qualified for--expired in April 2010 for loans closing through June 2010. By targeting first-time homebuyers, the program likely stimulated demand for high-LTV home-purchase mortgages. An FHA loan may have had particular appeal for such borrowers because the FHA allowed the tax credit to be used in advance as part of the down payment. This factor may potentially explain the decline in nonconventional market share in the latter part of 2010. However, another factor may also have been in play. In April 2010, the FHA raised its upfront underwriting fee 0.5 percentage point.⁴⁷ The share of nonconventional loans in the home-purchase market peaked in April--well before the end of the first-time homebuyer tax credit program--and fell about 4 percentage points, remaining at that level through the end of the year. Notably, the share of nonconventional loans in the refinance market, which was unaffected by the tax credit program, peaked in May and declined about 4 percentage points thereafter.

In the sections that follow, we discuss the differential implications that these changes may have had for particular groups and the potential effects that proposed changes in the GSE and FHA limits may have on the marketplace.

Demographic Patterns

As discussed earlier, 2008 and 2009 were characterized by the increased roles of the GSEs and of the FHA, VA, FSA, and RHS programs. Such government-related lending continued at roughly the same levels in 2010. This section examines how government-related lending played out differently across borrower groups. We differentiate among borrowers by race and ethnicity, relative income (for both the neighborhood and the borrower), location (state), type of lender, and indicators of low-quality lending.

Changes in the share and number of home-purchase and refinance loans from 2006 to 2010 for different groups are shown (tables 14.A and 14.B). These data indicate different patterns for home-purchase lending compared with refinance lending. For example, the shares of home-purchase loans to black and Hispanic-white borrowers decreased from 2006 to 2009, but the decrease in these groups' shares of the refinance market was more severe and continued in 2010. Also, the share of refinance loans to borrowers with low or moderate incomes (LMIs) fell significantly from 2006 to 2010, while the share of home-purchase loans to such borrowers increased significantly. Most of this growth took place in 2008 and 2009 but was sustained in 2010, when the first-time homebuyer tax credit program was still in place. Notably, the share of home-purchase lending to LMI borrowers fell significantly in the second half of 2010, after the homebuyer program expired (data not shown in tables).

We also show trends in two metrics of loan quality that can be derived from the HMDA data--the percentage of loans with estimated front-end debt-payment-to-income (PTI) ratios exceeding 30 percent (a warning level in underwriting) and the percentage of loans for non-owner-occupied properties.⁴⁸ Both measures fell significantly over the sample period, although most of this decline had taken place before 2009. In 2010, patterns for these measures diverge as the incidence of high-PTI lending declines and that of non-owner-occupant lending increases for both home-purchase and refinance lending.

Some of the changes from 2006 to 2010 may reflect factors specific to certain geographic areas rather than factors specific to certain demographic groups. For instance, a decline in lending in California relative to the rest of the nation would tend to generate a relative decline in lending to Hispanic white borrowers because of the prevalence of this group in California. Indeed, the share of loans extended to residents of the sand states--Arizona, California, Florida, and Nevada--declined, particularly for refinance lending from 2006 to 2009, rebounding some in 2010. Nevertheless, even after controlling for differential trends in lending across markets--that is, removing overall market trends from the analysis--the racial and income trends described earlier mostly remain (data not shown in tables).

Borrowers of different demographic groups showed large differences in their propensity to use different types of loans, with significant changes from year to year. All groups showed substantial increases in their use of nonconventional loans from 2006 through 2009 (data derived from tables 14.A and 14.B).⁴⁹ Black and Hispanic-white borrowers, however, relied particularly heavily on these government programs, a trend that continued in 2010. In 2010, more than 80 percent of home-purchase loans and more than 40 percent of refinance loans to black borrowers were nonconventional. For Hispanic white borrowers in 2010, nearly three-fourths of their home-purchase loans and 25 percent of their refinance loans were nonconventional. In 2006, over 40 percent of home-purchase and refinance loans to both black and Hispanic-white borrowers were sold into the private securities market or sold to a nongovernment purchaser. By 2007, these shares had dropped significantly, and the GSE and portfolio shares of loans among these groups had grown. In 2008 and 2009, the share of home-purchase loans to black and Hispanic-white borrowers that were sold to the GSEs fell, while the share of refinance loans to both groups that were sold to the GSEs rose from 2007 through 2009 and remained flat in 2010.

Patterns of loan-type incidence for LMI borrowers and borrowers living in LMI census tracts are similar to those for black and Hispanic-white borrowers but are more muted. Loans to these borrowers were less likely to be sold on the nongovernment secondary market in 2006, and the shift toward nonconventional loans in 2008 and 2009 was not as large. The share of borrowers with income missing from their loan applications fell from 2006 through 2009 (more than one-half of these loans were sold into the private secondary market in 2006). The incidence of missing income for refinance loans actually rose in 2008 and 2009, likely the result of "streamlined" refinance programs.

In 2006 and 2007, nonconventional loans as well as GSE loans were significantly less likely than portfolio or private secondary-market loans to be classified as low quality by our measures--high PTI or non-owner occupant. However, by 2008, this lower incidence for high-PTI loans had largely disappeared.

Loans originated in the sand states in 2006 and 2007 were much more likely to be sold into the private secondary market than loans originated in other states. By 2008, differences in the disposition patterns between the sand states and the rest of the country had largely disappeared in the home-purchase market, perhaps in part because of changes in the FHA and GSE loan limits. However, in the refinance market, loans originated in the sand states in 2008 and 2009 were more likely to be purchased by the GSEs and less likely to be part of the nonconventional loan programs than loans in other states.

Back to section top

Loan-Size Limits

Before 2008, the National Housing Act, as amended in 1998 Mortgagee Letter 1998-28, required that FHA mortgage limits for one- to four-family homes be set at 95 percent of the median house price prevailing in an area (either county or MSA), subject to an overall national minimum and maximum.⁵⁰ Loans purchased by the GSEs were also subject to a limit, based on national median house prices, which was fixed at $417,000 for single-family homes in the continental United States from 2006 to 2008. The Congress authorized an increase in these limits as part of the Economic Stimulus Act, passed in February 2008; it did so again as part of the Housing and Economic Recovery Act (HERA), enacted in July 2008; and it did so once more as part of the American Recovery and Reinvestment Act, passed in February 2009.⁵¹

The new FHA and GSE limits have remained in place, with only modest variation, since early 2008. However, barring congressional action, national single-family home loan limits on both FHA and GSE lending are scheduled to fall from $729,750 to $625,500 on October 1, 2011. Both FHA and GSE loan limits in areas not subject to the national cap are scheduled to fall from the current 125 percent to 115 percent of the area's median house price, with GSE single-family loans still subject to a base limit of $417,000.

Analysis presented in a previous article concluded that the increased loan limits accounted for less than 10 percent of the growth in nonconventional lending in 2008 and an even smaller portion of the growth in GSE loan purchases.⁵² Here we examine what the effects of the limit changes scheduled for October 1, 2011, are likely to be, based on lending patterns observed in 2010.

Analysis released by the Department of Housing and Urban Development (HUD) suggests that 669 counties and county equivalents, predominantly located in high-cost areas on both coasts, will face changed FHA loan limits for one- to four-family homes as of October 1, 2011.⁵³ Similar analysis by the FHFA suggests that 250 counties and county equivalents will face changes in GSE limits.⁵⁴ These numbers are not fully set, and some disagreement remains as to what the final changes will be.⁵⁵ Nevertheless, we use the projected limit changes forecast by HUD and FHFA to identify lending in 2010 in potentially affected areas.⁵⁶

All of the counties facing changes in GSE limits are in high-cost areas where 2010 GSE and FHA limits are the same. For about one-half of these counties, the FHA and GSE limits are projected to be reduced by the same amount, and future borrowers seeking loans in size ranges affected by the limits would not be able to use either the FHA or GSE programs. In the remaining counties facing GSE limit changes, the FHA limits are projected to fall below the $417,000 GSE base limit for single-family homes. In these counties, borrowers with loan requests between $417,000 and the current limits will no longer have access to either the FHA or GSE programs. Borrowers who will no longer be eligible for FHA loans with requests below $417,000 in these counties and borrowers in counties facing only declines in their FHA limits will still meet GSE loan-size standards. Consequently, in our analysis, we divide 2010 loans into eight groups based on proposed GSE and FHA limit changes: loans in counties with projected GSE limit changes with loan sizes (1) below both the 2010 GSE/FHA and proposed 2011 FHA limits, (2) above the 2011 FHA limit but below the proposed 2011 GSE limit, (3) below the 2010 GSE/FHA limit but above the proposed 2011 GSE limit, or (4) above both the 2010 and proposed 2011 GSE/FHA limits (jumbo loans); loans in counties with projected FHA but not GSE limit changes with loan sizes (5) below both the 2010 and proposed FHA 2011 limits, (6) below the 2010 FHA limit but above the proposed 2011 FHA limit, or (7) above both the 2010 and proposed 2011 FHA limits; and loans (8) in counties with no changes in either the GSE or FHA loan limits.

Totals for first-lien owner-occupant one- to four-family home-purchase and refinance 2010 lending based on these groupings are presented (tables 15.Aand 15.B). Lending totals are shown for the market as a whole and for various demographic and other loan characteristics. For comparison purposes, we also give overall figures for jumbo loans--those with loan sizes above the GSE (and FHA) loan limits for their areas in 2010.

Overall, only 1.3 percent of the 2010 home-purchase loans (and 1.3 percent of refinance loans) fell into a size range that is currently eligible for both the FHA and GSE programs but will not be eligible for either program under the proposed limits (column 3). An additional 2.1 percent of 2010 home-purchase loans (and 2.4 percent of refinance loans) would potentially have been affected by the FHA changes in markets where GSE limits are unchanged (column 6) or FHA limits fall more than GSE limits (column 2).

However, within these ranges, the proposed changes likely would have had a significant effect--53.4 percent of the home-purchase loans and 59.6 percent of the refinance loans originated in 2010 in size ranges that would have exceeded the proposed 2011 GSE size limits were sold to the GSEs. For FHA loans, the effect is somewhat smaller but still significant--43.7 percent of the home-purchase loans with sizes eligible under 2010 limits but ineligible under 2011 limits in counties with no GSE changes were FHA or VA loans. For refinance loans, only 15.2 percent of loans meeting these criteria were FHA or VA loans.

Borrowers affected by FHA limit changes but with loan sizes under the GSE limits would appear to be likely to have the GSEs as a viable option if the changes are implemented (although lending standards for FHA loans differ from those for loans eligible for purchase by the GSEs in ways other than just loan size). In 2010, 35.0 percent of home-purchase borrowers and 58.8 percent of refinance borrowers falling into a size range affected by the proposed FHA changes in areas where GSE limits are unchanged had their loans sold to the GSEs.

It is more difficult to know what options will be available for borrowers no longer eligible under either the GSE or FHA programs. On the one hand, the overall share of national lending for loans that would be affected by changes in the GSE limit was considerably higher in 2010 than in 2008 and 2009 (bottom of tables 15.A and 15.B). However, it is about the same as the share in 2006 and 2007, before the limits were raised. These figures suggest that factors other than GSE (and FHA) loan limits affected the relative amount of lending taking place within these bands.

On the other hand, if the loans affected by the GSE (and FHA) changes had been forced into the jumbo market in 2010, this move would have resulted in a 50 percent increase in the size of the national home-purchase jumbo market and a 63 percent increase in that of the national refinance jumbo market.⁵⁷ Holding such loans on the portfolios of originating institutions would have meant an increase of over 20 percent in portfolioed loans for institutions serving the 250 counties where limits were changed. These numbers are substantial and suggest that at least some of these loans would not have been originated or would have been originated only at higher prices.

Examination of the demographic data shows that borrowers with loan sizes eligible under 2010 limits but ineligible under 2011 limits for both GSE and FHA limit changes were disproportionately Asian, lived outside LMI census tracts, had non-LMI incomes, and used independent mortgage banks, relative to the average borrower. More than one-half of such borrowers lived in the sand states, and none lived in the "rust states" of Illinois, Indiana, Michigan, Ohio, and Wisconsin (because none of the affected counties lie in the rust states). Such borrowers were also more likely than average to have a PTI ratio exceeding 30 percent.

Borrowers facing only FHA limit changes similarly were less likely than average to live in LMI census tracts or have LMI incomes, but unlike those affected by the GSE limits, such borrowers show geographic and racial distributions similar to the national averages.

Back to section top

Credit Circumstances in Neighborhood Stabilization Program Neighborhoods

Concerns about neighborhoods experiencing high levels of housing market distress have been a particular focus of public policy in recent years. This focus has been motivated by the belief that elevated levels of foreclosure and property abandonment can adversely affect not only those directly involved in the foreclosures, but also others in the surrounding neighborhood.⁵⁸ Such negative externalities or spillover effects may arise as foreclosed and often vacant properties attract vandalism and crime, and these units may be poorly maintained, casting a pall over the neighboring properties and adversely affecting their market values.⁵⁹ In the extreme, these spillover effects can help create a self-reinforcing downward spiral that can devastate the quality of life in an area.

To address the foreclosure problem, as part of the 2008 HERA, the Congress established and funded the Neighborhood Stabilization Program.⁶⁰ The NSP provides emergency assistance to state and local governments seeking to support neighborhoods experiencing high levels of property abandonment and foreclosure. To help ensure that funds are appropriately targeted, the monies are directed to households or individuals with incomes less than 120 percent of the broader area median income.

To bolster congressional efforts, in December 2010, the federal bank and savings institution regulatory agencies revised the regulations that implement the Community Reinvestment Act (CRA) to support the stabilization of communities hard hit by elevated foreclosures.⁶¹ In particular, the revised regulations encourage covered institutions to support the NSP. Under the CRA rules, lenders are encouraged to make loans and investments and provide services to support NSP activities to individuals and neighborhoods beyond the traditional focus of the CRA (specifically, individuals and neighborhoods classified as lower income). Allowing banking institutions to receive CRA consideration for activities in NSP-targeted neighborhoods provides additional incentives for these institutions to leverage government funds targeted to these areas and populations.

Under the NSP program, funds may be used in different ways, including for the purchase or rehabilitation of abandoned or foreclosed properties, the demolition of blighted structures, and the redevelopment of demolished or vacant properties. NSP funds can also be used to help homebuyers purchase properties. The NSP is a nationwide program, but participation requirements may differ across states and cities.

In deciding which neighborhoods to target, HUD relies on a statistical model that estimates which neighborhoods are likely to be experiencing high rates of foreclosure and mortgage delinquencies. Based on the outputs of this model, each census tract is given an NSP score ranging from 1 to 20. Scores are scaled so that each score point is given to 5 percent of the census tracts. Census tracts with NSP scores in the top quintile ("high-NSP tracts"), those with scores of 17 to 20, are eligible for aid. "Lower-NSP tracts," those with scores below 17, are not generally eligible for aid unless they are in states that have very few tracts with NSP scores above 17, in which case the state is permitted to use a lower-threshold NSP score for identifying areas eligible for NSP funds.

An evaluation of the effectiveness of the NSP program is beyond the scope of this article. Some of the interventions, like the changes to the CRA, are too new to evaluate, and others require more data than HMDA provides. However, the HMDA data can be used to assess mortgage activity across all areas scored for the NSP program. Because the NSP program has been in existence for a few years, it is possible that recent loan flows may have been affected by the program to some degree. Nevertheless, this analysis can highlight some of the potential challenges involved in aiding these communities.

Substantial differences between high-NSP tracts and lower-NSP tracts existed long before the recent difficulties in mortgage and housing markets emerged. In the 2000 census, high-NSP tracts were characterized by higher minority concentrations and lower relative-income levels than lower-NSP tracts (table 16). Similarly, lending activity in these tracts before the subprime crisis was notably different. Using 2005 as a reference point, lending in the high-NSP tracts was characterized by elevated rates of loan denial; larger incidences of higher-priced loans, piggyback loans, and non-owner-occupant lending; and smaller shares of lending by lenders subject to the CRA. New home buyers in areas with high-NSP tracts also tended to have lower credit scores than buyers in other areas.

Since 2005, lending activity in high-NSP tracts has fallen faster than in lower-NSP tracts. In 2005, more home-purchase loans were extended in high-NSP tracts than in tracts in any of the other NSP score quintiles. Since 2005, declines in home-purchase lending volumes have been particularly steep in high-NSP neighborhoods. In 2010, home-purchase lending in high-NSP tracts was down 75 percent from 2005 levels. This decline was much more rapid than that experienced in the other NSP quintiles. As a result, in 2010, fewer loans were originated in the high-NSP tracts than in any of the other NSP quintiles, a reversal of the pattern observed in 2005.

One potential reason for the steeper decline in home-purchase lending in the high-NSP neighborhoods is offered by the role of the sand states. House price declines have been particularly steep in these states, and previous HMDA analyses have shown that mortgage lending has fallen more steeply in these states since the height of the housing boom. Because the high-NSP tracts are more likely to be located in the sand states than lower-NSP tracts--in 2005, the sand states accounted for 71 percent of loans to high-NSP neighborhoods, a share double that of any of the other NSP quintiles--we would expect the more-rapid lending declines experienced by the sand states to result in a faster decline in lending to high-NSP tracts. However, the HMDA data reveal that lending volumes in high-NSP tracts located outside of the sand states actually fell slightly more (73 percent) than in the sand states (71 percent). This result suggests that the declines in lending volumes that are observed for the high-NSP tracts do not simply reflect geographic differences.⁶²

A second possible reason for the steeper declines in home-purchase lending in the high-NSP neighborhoods is the role of the non-owner-occupant lending in these neighborhoods. Again, using 2005 as the base year, the share of home-purchase lending backed by non-owner-occupied properties in high-NSP tracts (25 percent) was 6 percentage points higher than in any of the lower-NSP quintiles. Since non-owner-occupant lending has fallen more rapidly than lending for owner-occupied properties across the board (as noted earlier), this finding can help explain some of the more-rapid decline in the lending activity in high-NSP neighborhoods. Non-owner-occupant lending fell 83 percent in the high-NSP tracts between 2005 and 2010--a decline that was higher than that observed for overall home-purchase lending in the high-NSP tracts or in the lower-NSP quintiles over the same period. Nevertheless, when the analysis is limited to owner-occupant lending, home-purchase lending has still fallen substantially in high-NSP tracts (68 percent) and at a rate that is well above the declines in lower-NSP tracts.

This outcome suggests that the steeper decline in lending in high-NSP neighborhoods appears to be broadly based, in that it has not been limited to non-owner-occupant lending or lending in specific states or MSAs. Instead, the steeper decline appears to reflect a changing pattern of home-purchase activity by higher-income borrowers. Loans to lower-income borrowers declined less steeply between 2005 and 2010 in high-NSP tracts (31 percent) than in lower-NSP tracts (36 percent). This pattern is reversed for lending to higher-income borrowers. In high-NSP tracts, loans to higher-income borrowers were 84 percent lower than they had been in 2005. While lower-NSP tracts also experienced sharp contractions, the declines have been less severe. The percentage decline in the high-NSP tracts was 13 percentage points above the fourth NSP quintile and 35 percentage points higher than the declines in the first quintile. The patterns for loans to middle-income borrowers have also contracted more sharply in high-NSP tracts, though the sizes of the differences have not been as large.

This changing income pattern of homebuyers suggests a challenge that efforts like the NSP confront in attempting to stabilize neighborhoods. Not only has home-purchase lending declined more rapidly in the highly distressed neighborhoods identified, but also the composition of the borrowers taking out loans has shifted notably toward those with lower incomes. While the share of loans going to higher-income borrowers in the lower-NSP quintiles declined from 50 percent in 2005 to 43 percent in 2010, in high-NSP tracts the decline was much steeper, falling from 52 percent in 2005 to 29 percent in 2010. This outcome suggests that much of the decline in lending in the highly distressed tracts reflects reduced inflows from higher-income borrowers. The lower income levels of new borrowers in the high-NSP tracts may inhibit the stabilization of these communities.

Back to section top