Excerpts from Fed Study on Bias in Home Lending
Since 1976, most banks and other depository institutions that have offices in metropolitan areas have been required, under the Home Mortgage Disclosure Act (HMDA), to disclose to the public information about the geographic distribution of their loans for home purchase and home improvement.
The data have revealed wide variations in the number and dollar volume of loans approved across neighborhoods grouped by the income and race of residents. These variations, together with data from other sources, have raised questions about whether the efforts of lenders have been adequate to help meet the credit needs of the low-income and minority residents of their communities.
The variations in lending patterns also have generated controversy about whether lenders treat applicants for home loans fairly and on a racially nondiscriminatory basis.
Some people interpret the variations as evidence of illegal discrimination. Others suggest that the patterns are attributable to differences in the demand for housing and home loans among individuals and across neighborhoods, and that they reflect the application of legitimate credit standards by lenders as they review individual requests for home loans.
The thrift-bailout law of 1989 substantially broadened the amount of information that lenders must file with the Fed. In the past, covered institutions were required to disclose information only on loans they originated or purchased.
Now, in disclosure statements released to the public this month, lenders for the first time have reported on all home loan applications they received and their disposition, plus the race or national origin, gender, and annual income of the applicants. In addition, more lenders are now subject to the reporting requirements.
The changes in the act's requirements, as implemented by the Federal Reserve Board's Regulation C, will increase the usefulness of the HMDA data to community organizations, local governments, financial institutions, and others.
The expanded data will make it possible, for example, to review how lenders act on applications and are likely to stimulate dialogue between institutions and members of their communities. Observed differences in the number of applications received and loans extended to various groups and neighborhoods are likely to lead financial institutions to reexamine their marketing and community outreach efforts.
Differences in approval and denial rates among groups and neighborhoods revealed by the new data can be expected to raise questions about the adequacy and fairness of the home lending process. The data have important limitations, however, and care must be taken in drawing conclusions from observed lending patterns.
Foremost among these institutions is a lack of information about factors that are important in determining the creditworthiness of applicants and the adequacy of the collateral offered as security for their loans.
Without taking into account such information, one cannot determine whether individual applicants or applicants grouped by a common characteristic (such as race or gender) have been treated fairly.
Major use of the expanded HMDA data will be made by the agencies charged with ensuring that covered institutions comply with the fair lending laws (the Fair Housing and Equal Credit Opportunity acts) and the Community Reinvestment Act (CRA).
Because bank examiners have access to loan application files, they will be able to overcome most of the limitations of the HMDA data.
Patterns Emerge
Because the 1990 HMDA data have just been released, little is yet known about what the expanded data may reveal once they are thoroughly analyzed. This section takes a first look at some loan and application patterns discernible from the data.
Myriad levels of analyses are possible, particularly with respect to different geographic areas and different groupings of financial institutions.
The focus here is on nationwide totals and on some potential uses of the new and expanded data. In reviewing the nationwide data, it should be noted that the lending records of individual institutions may vary greatly, both from one another and from patterns for the nation as a whole, depending on their location, the types of applicants they serve, the types of loan products they offer, and their credit standards.
The statistics presented here are based on preliminary data and are subject to revision. It is anticipated that revised data will be available in January 1992.
Volumes of Applications: In 1990, lenders covered by HMDA took action on roughly 5.26 million home loan applications - 3.09 million for purchase, 1.02 million for refinancing, and 1.10 million for improvement of residencies housing one to four families, and the balance for loans on multifamily dwellings for five or more families.
Among home purchase loan applications, 74% were for conventional mortgage loans and the remainder were for government-backed forms of credit - Federal Housing Administration (FHA), Veterans Administration (VA), and Farmers Home Administration (FmHA) loans.
Use of Various Home Purchase Loan Products: Application patterns for various kinds of home purchase loans differ according to applicant income. Government-backed loans are much more likely to be used by households with relatively low incomes than by households with high incomes.
The 1990 HMDA data indicate that 39% of applicants with low incomes (less that 80% of the median family income for their MSA) applied for government-backed home purchase loans, compared with only 15.6% of applicants with high incomes (more than 120% of the median family income for their MSA).
The new data also indicate that black (and to a much lesser extent Hispanic) applicants are more likely than either white or Asian applicants to seek government-backed home purchase loans.
Blacks in particular are relatively more likely to seek FHA and VA loans: Blacks constituted 4.3% of all applicants for conventional home purchase loans in 1990, but they accounted for 10.5% of all applicants for FHA loans and 11.7% of all applicants for VA credit.
Viewed in another way, 46% of all black home loan applicants applied for either an FHA or a VA loan, while only 28.6% of Hispanic applicants, 24.4% of white applicants, and 10.2% of Asian applicants sought such loans.
These simple summary statistics, though revealing, do not take into account the financial circumstances of the applicants that make up the various racial and ethnic groups. Income is the only financial characteristic of the applicant reported in the HMDA data.
After controlling for applicant income, however, the 1990 HMDA data still indicate that blacks, and to a lesser extent Hispanics, are more likely than whites to use FHA and VA loans. For instance, 60% of low-income black applicants sought government-backed home purchase loans, compared with 37% of low-income white applicants.
Overall Approval Rates: Lenders approved the majority of home purchase loan applications they received - roughly 72.3% of applications for conventional loans and 71.7% for government-backed loans.
Among the applications for conventional loans, 16.6% were denied by the lender and 10.2% were withdrawn by the consumer; in a relatively small number of cases (less than 1%) the application file was closed after the applicant was asked for but failed to submit information required for the credit decision. For government-backed home purchase loans, the denial rate was 16.5% and the withdrawal rate 10.6%.
The relatively high approval rates for home purchase loans likely reflect two characteristics of this market. First, prospective homebuyers frequently work with real estate sales agents who help them determine in advance of any application the size of the loan for which they are likely to qualify.
Second, because consumers incur up-front costs to file a home loan application - to cover, at a minimum, a property appraisal and credit bureau check - they have a strong incentive to learn about the prevailing standards for credit used by the industry and by particular lenders they might approach for credit.
Approval Rates for Minorities: Although the majority of home purchase loan applications are approved, many are not. Approval rates vary according to the applicant's income and demographic characteristics and the characteristics of the area in which the applicant resides or seeks to purchase a home.
Data previously available from sources other than HMDA indicate that blacks and Hispanics applying for mortgage loans at thrift institutions are significantly more likely than white applicants to be denied credit and that the experience of Asians is not greatly different from that of whites. The 1990 HMDA data reveal a similar pattern for all lenders covered by HMDA.
Conventional Home Purchase Loans: Nationally, about 14.4% of white applicants for conventional home purchase loans were denied credit in 1990. In sharp contrast, the rate for black applicants was 33.9% and for Hispanics 21.4%. At 12.9%, the denial rate for applicants of Asian extraction was lower than for any other racial or ethnic group.
Applicant income can be expected to affect the ability to qualify for a home purchase loan, but income is just one criterion considered by lenders in evaluations of creditworthiness. A household with a relatively low income may qualify for a loan of a given size and set of terms when a high-income household cannot because of differences in such things as level of their nonhousing debt, assets available upon down payment, employment experience, and credit history.
On average, however, low-income households have relatively fewer assets and lower net worth, experience more frequent employment disruptions, and are more likely than high-in-come households to fall behind on scheduled debt repayments.
The 1990 HMDA data reveal that the lower the income, the lower the acceptance rate. Nationwide, 78.9% of the loan applicants whose income equaled or exceeded the median family income for their MSAs were approved for conventional home purchase loans, compared with 69.4% of the loan applicants with lower incomes. Differences are even greater when comparisons are made at the extremes of the income distributions.
The national level of denial rates for applicants categorized by race or national origin reflects, in part, differences in the proportion of each group that has relatively low incomes. For example, among white applicants for conventional home purchase loans, 14% had incomes below 80% of their MSA'a median family income.
Low-income black and Hispanic applicants, in contrast, accounted for 25% and 16% of all applicants in their respective groups. Low-income Asians accounted for only 8% of the conventional home purchase loan application filed by Asians overall.
The differences in denial rates when applicants are grouped by race or national origin do not change notably when they also are categorized by income.
For example, among applicants whose incomes place them in the lowest income group, the denial rates for blacks, Hispanics and Asians were 40.1%, 31.1%, and 17.2%, respectively, compared with 23.1% for white applicants. Among applicants in the highest income group, denial rates for blacks, Hispanics, and Asians were 21.4%, 15.8%, and 11.2%, respectively, compared with 8.5% for whites.
The application withdrawal rate for conventional home purchase loans for both black and white applicants was 9.4%. The rates were higher for both Hispanics and Asian applicants, 12.4% and 13.5%, respectively.
The 1990 HMDA data also indicate some differenced when home loan applications are categorized by gender - male (one or more males), female (one or more females), or joint (one male and one female).
For instance, joint applicants are more likely than either male or female applicants to have a conventional home purchase loan approved. Female applicants are somewhat more likely than the male applicants to have a home purchase loan approved.
Government-Backed Home Purchase Loans: The pattern for denial for government-backed home purchase loans is similar to that for conventional home purchase loans. The rates of denial were 26.3% for blacks, 18.4% for Hispanics, and 12.8% for Asians, compared with 12.1% for whites. The rates of application withdrawal were 11.3% for blacks, 11.6% for both Hispanics and Asians, and 9.7% for whites.
Looking at disposition of applications for government-backed loans by gender, joint applicants are somewhat more likely than either male or female applicants to have a home purchase loan approved. Female applicants are more likely than male applicants to have a home purchase loan approved.
Home Improvement Loans: The patterns for denial and withdrawal of home improvement loan applications are broadly similar to those for home purchase applications. Generally, for all groups the denial rates are higher than for home purchase loans, and the withdrawal rates lower; 36.9% of black, 32.5% of Hispanic, and 24.6% of Asian applicants were denied loans, compared with 17% of white applicants.
Looking at disposition by gender, joint applicants were more likely than either male or female applicants to have a home improvement loan approved. Males are somewhat more likely than females to have a home improvement loan approved.
Relation of Approval Rates to Neighborhood Income and Composition: The HMDA data make it possible to compare lending across neighborhoods grouped by racial makeup and the income level of their residents. Considerable caution should be exercised, however, when making such comparisons. The usefulness of these data is currently limited by the lack of an up-to-date match with the characteristics of census tracts.
The recently released HMDA disclosure statements are based on 1980 census tract boundaries and population characteristics (neighborhood income level, racial composition, and housing stock characteristics). This census information is now more than 10 years old, and in some cases the resulting figures may be misleading.
For example, a low-income, predominantly minority neighborhood in 1980 may have undergone substantial change and may now have a much higher average income and a different racial composition. The Federal Reserve Board has published proposed amendments to HMDA reporting requirements, calling for a switch to the 1990 census tract definitions beginning in January 1992.
The Federal Financial Institutions Examination Council (FFIEC) plans to reflect socio-economic information about these areas in the disclosure tables portraying 1992 lending activity, which will be released in 1993.
Approval of Home Purchase Loan Application: Although the majority of applications for home purchase loans are approved, experience differs across neighborhoods grouped by racial composition and the income levels of their residents. The patterns of loan acceptance and denial do not differ greatly, whether the type of home purchase loan sought is conventional or government backed.
Neighborhood Income: The 1990 HMDA data indicate that the rate of loan denial declines as the income of the residents of an area increases. The rate of loan denial for conventional home loans relating to properties in low- or moderate-income neighborhoods was 20.2%, appreciably higher than the 13.9% for middle-income and 9.7% for upper-income neighborhoods.
For government-backed loans, the rates of loan denial were 17.8% for low- or moderate-income and 11.2% for upper-income neighborhoods.
Neighborhood Racial Composition: The 1990 HMDA data indicate that the rate of loan denial increases as the proportion of minority residents increases. For conventional home loans, the denial rate is about 12% for areas with less than 10% minority residents and rises to about 24% for areas with 80% or more minority residents. The pattern of loan denial for government-backed loans is virtually the same as that for conventional loans.
Neighborhood Income and Racial Composition: The difference in denial rates across neighborhoods of different racial composition is roughly the same even when differences in neighborhood median family income levels are taken into account.
For the most part, whether the neighborhood is low or moderate income, middle income, or upper income, the proportion of home purchase loan applicants denied credit increases as the percentage of minority residents increases. This pattern is present for applications for both conventional and government-backed forms of credit.
Approval of Home Improvement Loan Applications: Like home purchase loans, the majority of home improvement loan applications are approved regardless of neighborhood income or racial composition. Also like home purchase loans, the denial rate for home improvement loans increases as neighborhood income declines and the percentage of minority residents increases.
Limits of the Data
The HNDA data offer more detailed information about the home lending activities of reporting institutions, bringing the prospect for a better understanding of lending patterns through analyses previously not possible.
Knowing the personal characteristics of loan applications makes it feasible, for example to gauge more accurately the level of loan demand faced by an individual lender or a group of lenders seeking to serve different types of customers and various geographic areas within their communities. At the same time, the limitations of the data must be recognized.
The 1990 HMDA data document differences in the experiences of loan applicants grouped by their personal characteristics or by the characteristics of the neighborhoods in which they seek to purchase or improve homes.
Most prominently, the data indicate that black and Hispanic applicants are denied home loans more frequently than are white or Asian applicants who have similar incomes.
The data also indicate that applicants seeking to purchase homes in low- or moderate-income neighborhoods (regardless of the race of the residents) are denied credit more frequently than are applicants seeking to buy homes in upper-income neighborhoods.
The HMDA data can and should be used to raise questions about lending activity and to develop hypotheses for further investigation. The application-disposition patterns, however, reflect a wide variety of economic factors that determine the creditworthiness of individual home loan applicants and the adequacy of the collateral provided by the properties they seek to purchase or improve. Thus, caution in interpreting the numbers is called for.
For example, although the expanded HMDA data show loans denied by race or national origin, that information alone does not provide a basis for an independent assessment of whether an applicant who was denied credit was in fact creditworthy. Similarly, the HMDA data do not establish whether the property involved in the proposed credit extension was appropriately valued.
Thus, it is not possible to determine, from the HMDA data alone, whether loam applicants are being treated fairly and on a racially nondiscriminatory basis.
Fundamentally, the rates of approval and denial of loan applications reflected by the 1990 HMDA data represent the separate outcomes of a credit review process carried out by the more than 9,000 covered financial institutions located across the country.
That process seeks to ensure that individuals granted credit will repay their debt as scheduled and that, should they fail to do so, the collateral offered as security will pay off the loan plus costs associated with the foreclosure.
Consequently, lenders evaluate the factors that they believe allow them to predict an applicant's ability to repay; among these factors are several consumer financial characteristics - the proportion of the consumer's income that will need to be dedicated to the repayment of the proposed loan plus other outstanding debts, the level of equity (through the down payment) that the consumer is able and willing to put into the property, the consumer's employment experience and prospects, and the consumer's history of repaying debts.
Lenders also consider the appraised value of the property serving as collateral for the loan.
The HMDA data reveal little about financial characteristics of loan applicants - only their annual income. Even here, two applicants who have similar income may be strikingly different in their asset levels, existing debt burdens, and credit histories. Applicants of different race and gender may differ systematically in their financial characteristics.
Other sources of information, such as consumer surveys conducted by the Federal Reserve, provide extensive data on the financial situations of households grouped, for example, by annual income, race, or gender. Here, too, caution is called for.
Consumer surveys generally represent a wider population of respondents than do the HMDA data, which represent only individuals who have applied for a home loan.
To the extent that group profiles developed from these surveys reflect the characteristics of home loan applicants, such information may prove helpful in understanding variations in loan disposition rates among applicants grouped by race or gender.
Federal Reserve and other consumer surveys show the financial situation of households grouped by income. These data indicate that, compared with high-income households, low-income households tend to have relatively few assets available for a down payment on a home; they have consumer debt, tend to have relatively high repayment burdens, and are more likely to have fallen behind in their scheduled debt repayments; and generally have more periods of involuntary unemployment or reduced work hours.
Generally, black and Hispanic households are much more likely to be in a low-income grouping than are white households. For example, the median income of households headed by black and Hispanics is roughly 57% and 71% respectively of the median income of families headed by whites.
These disparities reflect, among other things, sharp differences in employment experiences. For example, in mid-1991 the national unemployment rate for blacks was nearly twice that of whites.
Also, the financial asset and net worth positions of nonwhite and Hispanic households are substantially different from those of whites. For instance, in 1986 the mean amount of financial assets held by black families was $5,900, compared with $64,000 for white families.
Differences in net worth were even more pronounced, with black families having an average net worth of $29,000 and white families $165,000.
Uses for New Data
Users of the HMDA data include community-based and other types of consumer interest organizations, financial institutions, state and local government agencies, and federal supervisory agencies. Community-based organizations have long used HMDA data in assessing the home lending activities of institutions in their communities.
Financial institutions covered by HMDA use the information: to evaluate the success of their loan marketing efforts and community outreach programs and to compare their performance with the home lending activities of their competitors. State and local governments find the data useful in identifying areas that may need assistance.
Supervisory agencies will be a major user of the expanded HMDA data. The new information will help them better assess the performance of financial institutions in satisfying their obligations under the Community Reinvestment Act and their compliance with the fair lending laws.
The expanded data HMDA enhance the agencies' ability to conduct the portion of CRA evaluations focusing on home lending.
For instance, in the past it was difficult to determine whether the geographic distribution of a lender's home purchase credit extensions reflected the demand for its loan products.
Although information about applications has been available to examiners, until now it has been available only through the original applications and loan documents. With ready access to a listing of applications from the Loan/Application Region (LAR) data, examiners will be able to identify easily the geographic distribution of a lender's loan applications.
Examiners can compare an institution's record with the records of other lenders serving the same locality to see if, for example, performance reflects an absence or low level of lending activity in the locality.
If some peer lenders are receiving a significant number of applications and are extending home loans, the data likely will focus greater attention on the institution's efforts to determine community credit needs, on its marketing and outreach programs, and on the mix of loan products it offers.
On the other hand, if peer lenders are receiving few applications for home loans, weak demand may be the explanation. Few applications might also indicate, however, that outreach efforts and marketing among all lenders are either ineffective or not aimed at the community in question.
The new HMDA data also can be used in assessing whether a lender has established a reasonable CRA community delineation.
Supervisory agencies also will use the expanded HMDA data in evaluating compliance with the fair lending laws - the Fair Housing Act and the Equal Credit Opportunity Act.
With the new information about applicant race or national origin, gender, and annual income, examiners will be able to look for statistical indicators of possible discrimination, such as differences in denial rates among groups. They will then review individual home loan application records for specific evidence of any disparate treatment.
Although different denial rates for majority and minority group applicants, for example, ultimately may be found to have a legitimate basis, the identification of such differences is one step in the assessment process.
Loan Sale Data
The 1989 amendments to HMDA require lenders to report the type of secondary market purchaser of home loans they sold during the year. The legislative history of the amendments indicates that Congress sought the new information to help identify, indirectly, secondary market requirements that might have a discriminatory effect on protected groups.
The HMDA data provide an opportunity for the first time to profile, for loans covered by HMDA, the characteristics of both the borrowers whose loans are purchased by secondary market entities and the neighborhoods in which they reside.
Because not all financial institutions that deal with secondary market institutions are covered by HMDA, the patterns revealed by the HMDA data may differ from those that would be observed in a review of all secondary market activity.
Moreover, information on borrower characteristics is not available for all loans sold by covered lenders - only for loans they themselves originated.
Although HMDA information about the census tract location of properties is available for roughly 75% of the loans sold to, or securitized by, secondary market entities, information on borrowers' race or national origin, gender, and income is available for only about two-thuds of the loans.
In most instances when information is unavailable, lenders had purchased the loans from other institutions and were not required to report applicant characteristics.
Participants in the secondary mortgage market buy and sell mortgage loans or securities backed by mortgage loans. They also guarantee payments on pass-through securities issued against pools of residential mortgage loans.
In so doing, they enable institutions that originate loans to raise new funds. By selling assets that are otherwise relatively illiquid, loan originators are able to extend additional loans or to use the funds in other ways.
Three government-sponsored agencies dominate secondary market activity - the Federal National Mortgage Association (FNMA, or Fannie Mae), the Federal Home Loan Mortgage Corporation (FHLMC, or Fredie Mac), and the Government National Mortgage Association (GNMA, or Ginnie Mae) - although banks, thrift institutions, insurance companies, and other entities are active as well. FNMA and FHLMC mainly buy conventional mortgage loans. Most of these loans are packaged into securities and sold to investors. GNMA does not purchase loans, but rather guarantees the timely payment of principal and interest for privately issued securities backed by FHA-insured and VA-guaranteed loans.
Secondary market institutions generally do not originate loans, but they do specify the underwriting guidelines that loans must meet to be eligible for purchase or securitization by the secondary market.
These guidelines and related loan-size purchase limitations vary among secondary market institutions; thus, it should be expected that, for the loans these institutions purchase or securitize, the characteristics of the borrowers and neighborhoods where properties are located will differ as well.
For example, in 1990 the FNMA and FHLMC limit on home purchase loans on single-family properties they purchased or securitized was $187,450. The maximum loan amounts backed by FHA insurance - between $67,500 and $124,875 (the larger amount corresponding to localities where housing costs were higher) - were the limits for GNMA's FHA-related activities.
The limit on VA loans eligible for the loan pools that GNMA would back was $144,0000 at the beginning of 1990, and was increased to $184,0000 during the year.
Other secondary market purchasers do not necessarily follow these loan-size limitations. In particular, so-called jumbo loans (those exceeding the loan limit set by FNMA and FHLMC) are purchased by depository institutions, pension funds, insurance companies, and others.
Basic underwriting guidelines (such as maximum loan-to-value ratios and monthly debt-to-income ratios) also differ among the secondary market participants, although FNMA and FHLMC follow essentially the same guidelines.
In the case of GNMA, underwriting standards are established by HUD and the VA. Given that HUD and the VA impose less stringent loan standards than originators of conventional loans, and that they have different rules about the size of loans they will back, it should be expected that, overall, FHA and VA borrowers will differ markedly from conventional loan users.
Consequently, borrowers whose loans are securitized by GNMA are also likely to differ from those whose loans are sold to or securitized by FNMA or FHLMC.
Borrowers using loans backed by GNMA may differ from those using loans supported by FNMA and FHLMC for another reason. FHA and VA loans are almost exclusively fixed-rate loans, whereas adjustable-rate mortgage loans (ARMs) are widely used in the marketplace (in 1990, ARMs accounted for about 30% of all loan originations).
Both FNMA and FHLMC buy and securitize many ARMs. Thus, it should be anticipated that differences among groups of borrowers who choose ARMs and those who choose fixed-rate loans will be reflected in sales to secondary market institutions as well.
Lenders covered by HMDA sold roughly 2.3 million loans to secondary market institutions in 1990. Most of the activity (some 70%) was with FNMA, FHLMC, and GNMA.
Not surprisingly, given GNMA's focus on government-backed loans, the HMDA data indicate that GNMA is supporting home purchase loans made to low- or moderate-income, and to a lesser extent minority households, relatively more often than are other secondary market institutions.
Overall, 22% of the loans backed by GNMA guarantees were made to families whose incomes were 80% or less of the median family income of the MSAs in which they reside.
The comparable figures for both FNMA and FHLMC were roughly 10%. The average 1990 income of borrowers whose loans were guaranteed by GNMA was $43,535, compared with $64,390 for FNMA and $63,914 for FHLMC.
Differences in borrower income are also reflected in the size of loans purchased or backed by secondary market institutions. In 1990, the average loan backed by GNMA was $73,730, compared with $101,050 for FNMA and $100,890 for FHLMC.
Compared with other secondary market purchasers, relatively more GNMA-supported borrowers purchased properties in low- and moderate-income and middle-income areas. [Tabular Data Omitted]
