Board of Governors of the Federal Reserve System

Neil Bhutta, Steven Laufer, and Daniel R. Ringo, of the Division of Research and Statistics, prepared this article. Jimmy Kelliher provided research assistance.

This article provides an overview of residential mortgage lending in 2016 and discusses a number of changes in mortgage market activity over time based on data reported under the Home Mortgage Disclosure Act of 1975 (HMDA). Mortgage debt is by far the largest component of household debt in the United States, and mortgage transactions can have important implications for households' financial well-being. The HMDA data are the most comprehensive source of publicly available information on the U.S. mortgage market, providing unique details on how much mortgage credit gets extended each year, who obtains this credit, and which institutions provide this credit.

HMDA requires most mortgage lending institutions with offices in metropolitan areas to disclose to the public detailed information about their home-lending activity each year. The HMDA data include the disposition of each application for mortgage credit; the type, purpose, and characteristics of each home mortgage that lenders originate or purchase during the calendar year; the census-tract designations of the properties related to those loans; loan pricing information; personal demographic and other information about loan applicants, including their race or ethnicity and income; and information about loan sales (see appendix A for a full list of items reported under HMDA).¹

HMDA was enacted to help members of the public determine whether financial institutions are serving the housing needs of their local communities and treating borrowers and loan applicants fairly, to provide information that could facilitate the efforts of public entities to distribute funds to local communities for the purpose of attracting private investment, and to help households decide where they may want to deposit their savings.² The data have proven to be valuable for research and are often used in public policy deliberations related to the mortgage market.³

Mortgage lending during 2016 occurred in the context of rising house prices, the continuation of an upward trend in prices evident since 2012.⁴ Mortgage interest rates remained low for most of the year, hovering just slightly above their historical lows reached in late 2012 and early 2013. Mortgage rates jumped sharply, however, following the November elections. Mortgage credit conditions continued to slowly ease, but credit remained more difficult to obtain for individuals with lower credit scores or hard-to-document incomes. According to the Senior Loan Officer Opinion Survey on Bank Lending Practices, much of the easing in mortgage underwriting that occurred over the course of 2016 was for loans that were eligible for purchase by the government-sponsored enterprises (GSEs).⁵ Growth in new housing construction remained sluggish despite the gains in house prices and strengthening demand for both new and existing homes.⁶

This article presents findings from the HMDA data describing mortgage market activity and lending patterns over time, including the incidence of higher-priced lending and rates of denial on mortgage applications, across different demographic groups and lender types.⁷ Some of the key findings are as follows:

1. The number of mortgage originations in 2016 rose 13 percent, to 8.4 million from 7.4 million in 2015. For loans secured by one- to four-family properties, growth was strong in both home-purchase originations--which increased to 4.0 million from 3.7 million in 2015--and refinance originations--which increased to 3.8 million from 3.2 million in 2015.
2. Black and Hispanic white borrowers increased their share of home-purchase loans for one- to four-family, owner-occupied, site-built properties in 2016, the third consecutive annual rise for both groups. The HMDA data indicate that 6.0 percent of such loans went to black borrowers, up from 5.5 percent in 2015, while 8.8 percent went to Hispanic white borrowers, up from 8.3 percent in 2015. The share of home-purchase loans to low- or moderate-income (LMI) borrowers decreased to 26 percent in 2016 from 28 percent in 2015.
3. The average value of home-purchase loans rose 3.2 percent in 2016, to $257,000, with similar increases for loans made to borrowers of different racial and ethnic groups. The average value of home-purchase loans to Hispanic white borrowers remained well below the 2006 peak, while the averages for Asian, black, and non-Hispanic white borrowers were all above their 2006-07 peaks.
4. Black and Hispanic white borrowers continued to be much more likely to use nonconventional loans (that is, loans with mortgage insurance from the Federal Housing Administration (FHA) or guarantees from the Department of Veterans Affairs (VA), the Farm Service Agency (FSA), or the Rural Housing Service (RHS)) than conventional loans compared with other racial and ethnic groups. In 2016, among home-purchase borrowers, 69 percent of blacks and 60 percent of Hispanic whites took out a nonconventional loan, whereas about 35 percent of non-Hispanic whites and just 16 percent of Asians did so.
5. The share of mortgages originated by nondepository, independent mortgage companies has increased sharply in recent years. In 2016, this group of lenders accounted for 53 percent of first-lien owner-occupant home-purchase loans, up from 50 percent in 2015.⁸ Independent mortgage companies also originated 52 percent of first-lien owner-occupant refinance loans, an increase from 48 percent in 2015. For the first time since at least 1995, nondepository, independent mortgage companies accounted for a majority of each of these types of loans.

Mortgage Applications and Originations

In 2016, 6,762 financial institutions--banks, savings associations, credit unions, and nondepository mortgage lenders--reported data under HMDA on the nearly 14 million home mortgage applications they received (including about 2.4 million applications that were closed by the lender for incompleteness or were withdrawn by the applicant before a decision was made), which resulted in about 8.4 million originations. The number of originations in 2016 was up from 7.4 million originations in 2015 (table 1).

Refinance mortgages for one- to four-family properties increased by 530,000, or 16 percent, from 2015 to 2016. One- to four-family home-purchase originations grew by 384,000, or 10 percent, from 2015. Most one- to four-family home-purchase loans are first liens for owner-occupied properties. In the past five years, the number of such loans grew over 70 percent, from less than 2.1 million in 2011 to 3.5 million in 2016. However, the number of such home-purchase originations remained well below its peak in 2005 and was near levels observed in the mid-1990s (figure 1).⁹ The number of first-lien home-purchase loans for non-owner-occupied properties--which are primarily used as rental properties and second homes--increased 8 percent, from 403,000 in 2015 to 436,000 in 2016.

In table 1, the volume of first-lien lending for owner-occupied properties is further disaggregated by loan and property type. (Versions of table 1 containing loan counts and dollar values by month are available in the Excel file posted online with this article.)¹⁰ In addition to lien and occupancy status, the HMDA data provide details on the type of property securing the loan (site-built or manufactured home) and on the type of loan (conventional or not).¹¹ As noted earlier, nonconventional lending involves loans with mortgage insurance or other guarantees from federal government agencies, including the FHA, the VA, the RHS, and the FSA. Conventional lending encompasses all other loans, including those sold to the GSEs Fannie Mae and Freddie Mac as well as those held in banks' portfolios.

Table 1. Applications and originations, 2004-16

Nonconventional loans usually involve high loan-to-value (LTV) ratios--that is, the borrowers provide relatively small down payments. For site-built properties, nonconventional home-purchase loans increased about 9 percent in 2016, while conventional loans rose about 12 percent. The nonconventional share of first-lien home-purchase loans for one- to four-family, owner-occupied, site-built properties stood at about 39 percent in 2016, little changed from 2015 and down from a peak of 54 percent in 2009 (figure 2).¹²

Figure 2 shows that the marked decline in the nonconventional share since 2009 reflects a decrease in the FHA share of loans, while the VA and FSA/RHS shares have been steadier. One factor that appears to help explain the fluctuations in the FHA share concerns changes in the up-front and annual mortgage insurance premiums (MIPs) that the FHA charges borrowers. For example, between October 2010 and April 2013, the annual MIP for a typical home-purchase loan more than doubled, from 0.55 percent of the loan amount to 1.35 percent.¹³ Drops in the FHA's market share have been observed each time the FHA has raised premiums. In January 2015, the annual MIP was reduced to 0.85 percent for most borrowers, and the FHA share of home-purchase loans subsequently increased.¹⁴

The remainder of table 1 provides additional details on the breakdown of one- to four-family home-purchase and refinance loans by lien and occupancy status and by property and loan type.¹⁵ Table 1 also provides the number of applications for and originations of home-improvement loans for one- to four-family properties, many of which are junior liens or unsecured, and loans for the purchase of multifamily properties (consisting of five or more units). Finally, the HMDA data include details about preapproval requests for home-purchase loans and loans purchased by reporting institutions during the reporting year, although the purchased loans may have been originated at any point in time. Lenders reported roughly 530,000 preapproval requests; roughly 67 percent of these requests turned into an actual loan application for a specific property in 2016.¹⁶ Table 1 also shows that, for 2016, lenders purchased 2.2 million loans from other institutions.

The HMDA Data's Coverage of the Mortgage Market

It is important to note that the HMDA data do not provide universal coverage of residential mortgage lending in the United States. There are two main reasons HMDA coverage is not universal. First, not all lenders are required to report data. Among deposit-taking institutions like banks, the smallest institutions as well as institutions without any branches in a metropolitan statistical area (MSA) do not have to report data. Among institutions that take no deposits, nonprofits, smaller institutions, and those that operate entirely outside of an MSA also do not have to report data.¹⁷

Second, not all types of mortgage originations are reported. In particular, lenders do not report mortgages that are not for the purpose of purchasing a residential property, refinancing an outstanding mortgage, or making home improvements. Thus, a mortgage taken out solely to finance education expenses, for example, would not be reported. In addition, home equity lines of credit (HELOCs), regardless of their purpose, are not required to be reported under current rules.¹⁸

One way to assess the coverage of the HMDA data is to compare the number of loans reported under HMDA with the number of loans reported in consumer credit files. In contrast to the HMDA data, all mortgage loans regardless of purpose can be reported in consumer credit files, and all financial institutions have an incentive to report their mortgage loans to consumer credit bureaus, since reporting encourages borrowers to make on-time payments.¹⁹ According to estimates based on the consumer credit records maintained by Equifax, one of the three nationwide consumer credit-reporting agencies, about 8.4 million first-lien home-purchase and refinance loans were originated during 2016, compared with nearly 7.7 million first-lien home-purchase and refinance loans for one- to four-family properties reported under HMDA.²⁰ Thus, the number of first-lien home-purchase and refinance loans in the HMDA data is approximately 90 percent of the number reported in consumer credit files.²¹

In addition, Equifax estimates that about 874,000 home equity loans and 1.4 million HELOCs were originated in 2016. Many of these loans may not be reported under HMDA for the reasons stated earlier. If they were reported under HMDA, they would be classified as home-improvement loans or as junior-lien home-purchase or junior-lien refinance loans. Lenders reported about 658,000 such loans in the 2016 HMDA data, less than 30 percent of the number of home equity loans and HELOCs reported by Equifax.

Mortgage Outcomes by Income and by Race and Ethnicity

A key attribute of the HMDA data is that they help policymakers and the broader public better understand the distribution of mortgage credit across different demographic groups. The next set of tables provides information on loan shares, product usage, denial rates and reasons, and mortgage pricing for population groups defined by applicant income, neighborhood income, and applicant race and ethnicity (tables 2-8). With the exception of table 8, which includes loans for manufactured homes (and contains information by type of loan rather than by applicant or neighborhood characteristic), these tables focus on first-lien home-purchase and refinance loans for one- to four-family, owner-occupied, site-built properties. Such loans accounted for about 81 percent of all HMDA originations in 2016.

The Distribution of Home Loans across Demographic Groups

Table 2 shows different groups' shares of home-purchase and refinance loans and how these shares have changed over time. For example, black borrowers' share of home-purchase loans (conventional and nonconventional loans combined) was 6.0 percent in 2016, up from 5.5 percent in 2015. Similarly, the Hispanic white share of home-purchase loans was 8.8 percent in 2016, up from 8.3 percent in 2015. Both shares remain well below their 2006 peaks (8.7 percent and 11.7 percent for black and Hispanic white borrowers, respectively) but have now increased for two years in a row.²²

Table 2. Distribution of home loans, by purpose of loan, 2004-16

In terms of borrower income, the share of home-purchase loans to LMI borrowers slipped from 28.0 percent in 2015 to 26.2 percent in 2016.²³ In accordance with definitions used by the federal bank supervisory agencies to enforce the Community Reinvestment Act, LMI borrowers are defined as those with incomes of less than 80 percent of estimated current area median family income (AMFI); AMFI is estimated based on the incomes of residents of the metropolitan area or nonmetropolitan portion of the state in which the loan-securing property is located.²⁴

The HMDA data also shed light on borrowing patterns across neighborhoods, defined as census tracts. From 2015 to 2016, the share of home-purchase loans originated in high-income census tracts decreased slightly from 41.0 percent to 40.0 percent.²⁵ LMI and middle-income tracts both saw small gains. In table 2, it is important to note that shares by neighborhood income in 2012 and thereafter are not perfectly comparable with those in 2011 and earlier because census-tract definitions and census-tract median family income estimates were revised in 2012. The current tract demographic measures are based on 2010 census data and 2006-10 American Community Survey data, whereas the 2004-11 data relied on 2000 census income and population data.²⁶ In addition, the Office of Management and Budget published new metropolitan area delineations in 2014, so caution should be exercised in comparing relative income measurements between 2013 and later years.

Average Loan Size by Demographic Group

Table 3 shows the average dollar value of home-purchase and refinance loans by different groups and how these averages have changed over time. All dollar amounts are reported in nominal terms. The data reveal significant differences in the value of loans to different racial and ethnic groups. Asian borrowers took out the largest loans, averaging $373,000 for home purchases and $368,000 for refinancings in 2016, whereas loans to black borrowers averaged $217,000 for home purchases and $212,000 for refinancings.²⁷

Table 3. Average value of home loans, by purpose of loan, 2004-16

Over time, home-purchase dollar values have followed the historical trend of home prices, rising during the mid-2000s, falling sharply through 2008 and 2009, and then beginning to recover since about 2011. The trends differ substantially by race and ethnicity, however. The average home-purchase loan to a Hispanic white borrower in 2016 was for $220,000, up from $209,000 in 2015 but still well below the peak of $238,000 in 2006. In contrast, the average home-purchase loan amount for a non-Hispanic white borrower was about $246,000 in 2016, higher than the pre-crisis peak in 2007 of about $222,000. The values of loans to Asian and black borrowers in 2016 were also higher than their 2006-07 peaks.

In terms of borrower income, for LMI borrowers, the average home-purchase loan amount increased to $146,000 in 2016 from $141,000 in 2015; loan amounts increased by a similar magnitude for middle- and high-income borrowers. Average loan values also increased across all borrower-income groups for refinance loans.

Jumbo Lending

The share of conventional jumbo loans--those with loan amounts in excess of the GSEs' conforming loan limits and no other government guarantee--was about 5 percent in 2016, little changed from 2015. As shown in table 3, conventional jumbo loans made up 5.2 percent of all first-lien home-purchase loans for owner-occupied, one- to four-family, site-built homes in 2016, roughly the same as in 2015.²⁸ Among refinance loans, the conventional jumbo fraction decreased to 4.6 percent from 4.9 percent in 2015. Because of their larger size, jumbo loans make up a correspondingly larger share of the dollar volume of mortgages, accounting for 16.8 percent of home-purchase loans and 15.7 percent of refinance loans in 2016. Since the financial crisis, most new jumbo loans have been held on the originating bank's portfolio, as the market for mortgage-backed securities without a government guarantee is thin.²⁹

Variation across Demographic Groups in Nonconventional Loan Use

Table 4 shows that black and Hispanic white borrowers are much more likely to use nonconventional loans (FHA, VA, RHS, and FSA loans) than conventional loans compared with other racial and ethnic groups. In 2016, among home-purchase borrowers, about 69 percent of blacks and 60 percent of Hispanic whites took out a nonconventional loan, whereas about 35 percent of non-Hispanic whites and just 16 percent of Asians did so. These numbers have declined from their peaks in 2009 and 2010, when well over three-fourths of black and Hispanic white borrowers and over one-half of non-Hispanic white borrowers took out nonconventional loans.

Table 4. Nonconventional share of home loans, by purpose of loan, 2004-16

Nonconventional usage is also more prevalent for borrowers with lower incomes and in neighborhoods with lower incomes. In 2016, about 52 percent of LMI home-purchase borrowers and 49 percent of those borrowing to purchase homes in LMI neighborhoods used nonconventional loans, compared with about 27 percent of high-income borrowers and 28 percent of borrowers in high-income neighborhoods. While black and Hispanic white borrowers tend to have lower incomes, on average, than non-Hispanic white borrowers, the previously mentioned racial and ethnic differences in nonconventional loan use persist within income groups.³⁰ With respect to refinance loans, minority and lower-income borrowers are again more likely to use nonconventional than conventional loans. In general, however, nonconventional loans are less prevalent in refinance lending.³¹

Greater reliance on nonconventional loans may reflect the relatively low down-payment requirements of the FHA and VA lending programs, which serve the needs of borrowers who have few assets to meet down-payment and closing-cost requirements.³² The patterns of product incidence could also reflect the behavior of lenders to some extent; for example, concerns have been raised about the possibility that lenders steer borrowers in certain neighborhoods toward such loans.³³

Denial Rates and Denial Reasons

In 2016, the overall denial rate on applications for home-purchase loans was 11.4 percent, somewhat lower than in 2015 (table 5). The decrease in 2016 continues a trend of declining denial rates for home-purchase mortgages over the past decade.³⁴ In addition, denial rates have exhibited significant variation, and changes in denial rates have differed by type of loan. For example, for conventional home-purchase loan applications, the denial rate of 10.2 percent in 2016 was 8.3 percentage points lower than in 2006, while for nonconventional home-purchase loan applications, the denial rate of 13.4 percent in 2016 was 1.3 percentage points higher than in 2006. Variations in raw denial rates over time reflect not only changes in credit standards, but also changes in the demand for credit and in the composition of borrowers applying for mortgages. For example, the denial rate on applications for conventional home-purchase loans was lower in 2016 than during the housing boom years, even though most measures of credit availability suggest that credit standards are tighter today.³⁵ This result may stem from a relatively large drop in applications from riskier applicants.

Table 5. Denial rates, by purpose of loan, 2004-16

As in past years, black, Hispanic white, and "other minority" borrowers had notably higher denial rates in 2016 than non-Hispanic white borrowers, while denial rates for Asian borrowers were more similar to those for non-Hispanic white borrowers. For example, the denial rates for conventional home-purchase loans were about 22 percent for black borrowers, 15 percent for Hispanic white borrowers, 11 percent for Asian borrowers, 17 percent for other minority borrowers, and 8 percent for non-Hispanic white borrowers.

Previous research and experience gained in the fair lending enforcement process show that differences in denial rates and in the incidence of higher-priced lending (the topic of the next subsection) among racial or ethnic groups stem, at least in part, from factors related to credit risk that are not available in the HMDA data, such as credit history (including credit score), ratio of total debt service payments to income (DTI), and LTV ratio. Differential costs of loan origination and the local competitive environment, as well as illegal discrimination, may also bear on the differences in pricing.

Despite these limitations, the HMDA data play an important role in fair lending enforcement. The data are regularly used by bank examiners to facilitate the fair lending examination and enforcement processes. When examiners for the federal banking agencies evaluate an institution's fair lending risk, they analyze HMDA price data and loan application outcomes in conjunction with other information and risk factors that can be drawn directly from loan files or electronic records maintained by lenders, as directed by the Interagency Fair Lending Examination Procedures.³⁶ The availability of broader information allows the examiners to draw stronger conclusions about institution compliance with the fair lending laws.

Lenders can, but are not required to, report up to three reasons for denying a mortgage application, selecting from nine potential denial reasons (as shown in table 6). Among denied first-lien applications for one- to four-family, owner-occupied, site-built properties in 2016, about 72 percent of denied home-purchase applications and about 46 percent of denied refinance applications had at least one reported denial reason. The two most frequently cited denial reasons for both home-purchase and refinance loans were the applicant's credit history and DTI ratio (note that the sum across columns in table 6 can add up to more than 100 percent because lenders can cite more than one denial reason). For both home-purchase and refinance applications, the DTI ratio and collateral are more likely to be cited as denial reasons on conventional than nonconventional applications.

Table 6. Reasons for denial, by purpose of loan, 2016

Denial reasons vary across racial and ethnic groups to some degree. For example, among denied home-purchase loan applications in 2016, credit history was cited as a denial reason for about 24 percent of denied black applicants, 17 percent of denied Hispanic white applicants, 18 percent of denied non-Hispanic white applicants, and just 11 percent of denied Asian applicants. The DTI ratio was cited most often as a denial reason for home-purchase applicants in all racial and ethnic groups. For Asian home-purchase applicants, collateral was the second most common reason cited for denial, while for other groups, credit history was the second most common reason cited.

The Incidence of Higher-Priced Lending

Current price-reporting rules under HMDA, in effect since October 2009, define higher-priced first-lien loans as those with an annual percentage rate (APR) of at least 1.5 percentage points above the average prime offer rate (APOR) for loans of a similar type (for example, a 30-year fixed-rate mortgage).³⁷ The spread for junior-lien loans must be at least 3.5 percentage points for such loans to be considered higher priced. The APR of a mortgage differs from the interest rate because the APR measure incorporates up-front fees and other loan costs such as mortgage insurance. The APOR, which is published weekly by the Federal Financial Institutions Examination Council, 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.³⁸

In 2016, the fraction of home-purchase loans (again, first liens for one- to four-family, owner-occupied, site-built properties) above the higher-priced threshold was 7.7 percent (table 7.A). Although the overall fraction of higher-priced loans was little changed from 2015, there were moderate changes in the share of higher-priced lending among different types of loans. In particular, the higher-priced share of nonconventional loans declined from 14.5 percent to 13.9 percent, while the higher-priced share of conventional loans rose from 3.2 percent to 3.7 percent.

Table 7. Incidence of higher-priced lending, by purpose of loan, 2004-16

Table 7.A also shows that, in 2016 as well as earlier years, black and Hispanic white borrowers had the highest incidences of higher-priced loans within both the conventional and nonconventional loan types. This table provides the raw rates of higher-priced lending by group from 2004 to 2016, but, as discussed in detail in previous Bulletin articles, the raw rates reported in the public HMDA data can be difficult to compare over longer time horizons for two main reasons. First, a different price-reporting rule was in place prior to October 2009, with the spread between a mortgage's APR and the rate on a Treasury bond of comparable term (rather than the APOR) reported if it rose above 3 percentage points.³⁹ Second, even data from years prior to 2009 are not easily comparable from year to year, as the price-reporting rule used during this period created unintended distortions in reporting over time (which is why the reporting rule was changed).⁴⁰

Table 7.B provides adjusted rates of higher-priced lending that are intended to be more comparable over time. Using the dates of application and origination (which are not released in the public HMDA data files) and assuming all loans are 30-year fixed-rate mortgages, we can estimate the APR of loans that were originated when the old pricing rule was in effect.⁴¹ This estimated APR can then be compared with the APOR, as is done under the new price-reporting rule. Finally, because the implied threshold spread over the APOR during the previous reporting regime got to as high as about 2.5 percentage points, table 7.B reports the fraction of loans with an estimated APR spread over the APOR (or the actual reported spread for loans made under the new rules) of at least 2.5 percentage points--rather than 1.5 percentage points, as in table 7.A.⁴² Higher-priced lending by this measure virtually disappeared by 2008 and has not reemerged, likely reflecting the lack of subprime mortgage lending.

Table 7. Incidence of higher-priced lending, by purpose of loan, 2004-16

The higher-priced fraction of FHA home-purchase loans was about 21 percent in 2016, much higher than for other types of loans, in part because of the relatively high up-front and annual MIPs charged by the FHA (table 8). In contrast, only about 1 percent of VA/RHS/FSA home-purchase loans were higher priced. The higher-priced fractions for both types of loans were little changed from 2015.

Table 8. Distribution of price spread, 2016

HOEPA Loans

Under the Home Ownership and Equity Protection Act (HOEPA), certain types of mortgage loans that have interest rates or fees above specified levels are subject to additional consumer protections, such as special disclosures and restrictions on loan features. New rules extending HOEPA's protections from refinance and home equity loans to also include home-purchase loans and HELOCs became effective on January 10, 2014. These rules also added new protections for high-cost mortgages, such as a pre-loan counseling requirement for borrowers.

The new rules also changed the benchmark used to identify high-cost loans that are covered by HOEPA's protections. Instead of using the yield on Treasury securities, high-cost loans are identified by comparing a loan's APR with the APOR. HOEPA coverage now applies to first liens with an APR more than 6.5 percentage points above the APOR. If the loan is a junior lien or the loan amount is less than $50,000 and the loan is secured by personal property (such as a manufactured home), then the high-cost threshold is 8.5 percentage points above the APOR. Prior to 2014, HOEPA's protections were triggered if the loan's APR exceeded 8 percentage points above the rate on a Treasury security of similar term for first liens, and 10 percentage points for junior liens. Finally, under the new rules, HOEPA coverage is also triggered if the points and fees exceed certain thresholds.⁴³

While HOEPA loans were never a large fraction of the mortgage market, they have become even rarer since the housing boom. In 2005, lenders reported nearly 36,000 HOEPA loans (table 9). In 2016, the total was only 1,903 loans, despite the extension of HOEPA protections to home-purchase loans.⁴⁴

Table 9. Distribution of HOEPA loans, by characteristic of loan, 2004-16

Lending Institutions

In 2016, there were 6,762 reporting institutions (table 10). The total consisted of 3,805 banks and thrifts (hereafter, banks), of which 3,033 were small, defined as having assets of less than $1 billion; 1,939 credit unions; 131 mortgage companies affiliated with depositories (banks and credit unions); and 887 independent mortgage companies.⁴⁵ Banks collectively accounted for about 38 percent of all reported mortgage originations; independent mortgage companies, about 48 percent; credit unions, about 9 percent; and affiliates, the remainder. Over the past few years, the share of loans originated by independent mortgage companies has risen sharply. In 2016, these lenders originated 53 percent of first-lien owner-occupant home-purchase loans, up from 50 percent in 2015 and just 35 percent in 2010. Independent mortgage companies also originated 52 percent of first-lien owner-occupant refinance loans, an increase from 48 percent in 2015.

Table 10. Lending activity, by type of institution, 2016

Many institutions report little activity. About 43 percent of institutions (2,900 out of 6,762) reported fewer than 100 mortgage originations in 2016, accounting for about 118,000 originations, or less than 2 percent of all originations. About 15 percent of institutions originated fewer than 25 loans, in total accounting for about one-tenth of 1 percent of all originations.⁴⁶

Table 10 provides several other statistics to help compare the lending patterns of different types of institutions in 2016, and we discuss some highlights here. First, depositories tend to originate a significantly higher fraction of conventional loans than nondepositories. Second, small banks and credit unions have accounted for a highly disproportionate share of conventional higher-priced loans in recent years.

Third, independent mortgage companies originated a higher share of their home-purchase loans to minority borrowers and in LMI neighborhoods than other types of lenders. Fourth, large banks originate a significantly lower share of home-purchase mortgages to LMI borrowers compared with other types of lenders. In a supplementary analysis, we show that much of the decline in lending by large banks to LMI borrowers in recent years can be explained by their decreasing share of FHA-guaranteed mortgages, which are more likely to be originated to LMI borrowers.⁴⁷

Finally, the HMDA data provide information on whether originated loans were sold within the same calendar year and the type of institution to which they were sold, such as one of the GSEs or a banking institution (see appendix A for a full list of purchaser types). Table 10 displays the fraction of loans sold within the calendar year, as opposed to being held in portfolio.⁴⁸ Nondepositories sold virtually all of their loans in 2016. In contrast, credit unions sold about one-half of the home-purchase loans they originated and about 40 percent of the refinance loans they originated. That said, portfolio lending among depositories has declined significantly over the longer term.⁴⁹

Table 11 lists the top 25 reporting institutions according to their total number of originations, along with the same set of lending characteristics as those listed in table 10.⁵⁰ Wells Fargo reported the most originations, with about 438,000.⁵¹ The next-highest total was for Quicken Loans, followed by JPMorgan Chase (JPMC) and Bank of America. Overall, the top 25 lenders accounted for about 34 percent of all loan originations in 2016, up slightly from 33 percent in 2015. These same firms also provided additional funding by purchasing over 1.2 million loans from other lending institutions during 2016 (these loans could have been originated in 2015 or in earlier years), equal to about 43 percent of the number of loans they originated during the year.

Table 11. Top 25 respondents in terms of total originations, 2016

Table 11. Top 25 respondents in terms of total originations, 2016

The top institutions differ significantly in their lending patterns. Some of this variation reflects differences between types of institutions, which were discussed earlier. For example, table 11 shows that large banks like Bank of America have a higher share of conventional mortgages and a smaller share of lending in LMI neighborhoods compared with independent mortgage companies like Quicken Loans.

In addition to the variation across lender types, there is substantial variation in lending patterns within lender types. For example, regarding large banks, about 97 percent of JPMC's home-purchase loans were conventional, compared with about 82 percent for U.S. Bank. There is also significant variation among the large banks in the fraction of loans sold, with Wells Fargo selling almost 73 percent of its home-purchase originations, compared with less than 46 percent for Bank of America.

Finally, the composition of borrowers varied across the top 25 institutions, both within and across lender types. For some institutions, more than 30 percent of home-purchase borrowers were LMI, while at other institutions, fewer than 20 percent of borrowers were in that category.⁵² Although it is difficult to know precisely why such variation exists, these differences could reflect different business strategies or different customer demands in the markets and geographic regions the institutions serve, among other possibilities.

 

Appendix A: Requirements of Regulation C

Regulation C requires lenders to report the following information on home-purchase and home-improvement loans and on refinancings:⁵³

For each application or loan

• application date and the date an action was taken on the application

• action taken on the application

  • approved and originated
  • approved but not accepted by the applicant
  • denied (with the reasons for denial--voluntary for some lenders)
  • withdrawn by the applicant
  • file closed for incompleteness

• preapproval program status (for home-purchase loans only)

  • preapproval request denied by financial institution
  • preapproval request approved but not accepted by individual
• loan amount

• loan type

  • conventional
  • insured by the Federal Housing Administration
  • guaranteed by the Department of Veterans Affairs
  • backed by the Farm Service Agency or Rural Housing Service

• lien status

  • first lien
  • junior lien
  • unsecured

• loan purpose

  • home purchase
  • refinance
  • home improvement

• type of purchaser (if the lender subsequently sold the loan during the year)

  • Fannie Mae
  • Ginnie Mae
  • Freddie Mac
  • Farmer Mac
  • private securitization
  • commercial bank, savings bank, or savings association
  • life insurance company, credit union, mortgage bank, or finance company
  • affiliate institution
  • other type of purchaser

For each applicant or co-applicant

• race
• ethnicity
• sex
• income relied on in credit decision

For each property

• location, by state, county, metropolitan statistical area, and census tract

• type of structure

  • one- to four-family dwelling
  • manufactured home
  • multifamily property (dwelling with five or more units)
• occupancy status (owner occupied, non-owner occupied, or not applicable)

For loans subject to price reporting

• spread above comparable Treasury security for applications taken prior to October 1, 2009
• spread above average prime offer rate for applications taken on or after October 1, 2009

For loans subject to the Home Ownership and Equity Protection Act

• indicator of whether loan is subject to the Home Ownership and Equity Protection Act

Footnotes