CRA and fair lending regulations: resulting trends in mortgage lending.

Evanoff, Douglas D. ; Segal, Lewis M.

In response to concerns that banks were not adequately serving the credit needs of their local communities and not treating all applicants fairly, during the 1960s and 1970s Congress passed the fair lending laws and the Community Reinvestment Act (CRA).(1) These laws, aimed at eliminating discriminatory lending practices and encouraging lending to low-income individuals and in low-income areas, have been controversial since their inception. Community advocates argued that the acts were either inadequate or inadequately enforced and that banks continued to channel deposits away from local communities, resulting in inadequate financing for the areas most in need. Bankers argued that they treated applicants fairly and the acts smacked of credit allocation that could adversely affect bank safety and soundness.

Although there continues to be significant disagreement regarding these regulations, recently there has been a wave of positive reviews of their effectiveness.(2) The regulations have been given credit for encouraging banks to implement special loan programs aimed at lower-income communities and for effectively channeling funds toward previously underserved areas and minority groups. Community advocates argue that significant progress has been made and continued enforcement will reap additional benefits. Some bankers state that in responding to the CRA they have discovered new, profitable, previously untapped lending opportunities. These opportunities have come at a most convenient time as the demand in traditional lending markets has slowed.

While the arguments for the fair lending laws and the CRA are essentially ones of equity, there may also be economic arguments for constraining private market behavior and channeling funds to underserved areas. It may be that these credit flows produce positive externalities which, from a societal perspective, generate a total return greater than that received by the providers of the credit.(3) That is, although society reaps the full benefits of providing this credit, the service provider (a bank in this case) may not. While this provides economic justification for channeling credit to particular markets, it does not necessarily warrant doing so through the banking system.

In this article, we examine the evolution of the fair lending regulations and the CRA. We then summarize the economic literature that pertains to these regulations. Finally, we evaluate the effectiveness of the fair lending laws and the CRA by analyzing recent trends in mortgage lending activity and discussing whether these trends are in line with the intent of the regulations. We ask whether the trends can be attributed to the regulations and whether the data suggest that the regulations have been successful.

Evolution of the CRA and the fair lending laws

Although it is common to group together the Fair Housing Act, the Equal Credit Opportunity Act (ECOA), and the CRA, they are more accurately classified into two groups: the fair lending laws and the CRA. The fair lending laws are aimed at eliminating lending discrimination based on the inherent attributes of the borrower, such as race or gender. The CRA primarily addresses geographic discrimination, that is, failing to serve the credit needs of the local community in which the bank was chartered. The Home Mortgage Disclosure Act (HMDA) provides information on lending to individuals and locations, supporting the enforcement of both the fair lending laws and the CRA.

Fair lending laws

The Fair Housing Act was approved by Congress in 1968 as part of the Civil Rights Act of that year. It prohibits discrimination in residential real estate transactions based on race or color, religion, national origin, gender, handicap, or family status.(4) The ECOA encompasses a broader array of transactions. Passed in 1974, it prohibits discrimination with regard to any aspect of a credit transaction (consumer, commercial, or real estate loan) based on race or color, religion, ethnic origin, gender, marital status, age, and receipt of public assistance.(5)

It has been argued that fair lending enforcement prior to the 1990s was generally unaggressive.(6) The techniques employed to detect discrimination (reviewing whether internal policies were followed and performed uniformly across the protected factors) typically detected only the most blatant cases of discrimination. Since that time, in response to growing public concern about lending discrimination and well-publicized research that reported evidence of discrimination, regulatory agencies and the U.S. Department of Justice have stepped up their enforcement efforts. For example, a 1988 study of mortgage discrimination in Atlanta led the Justice Department to initiate an investigation into fair lending practices by depositories in that market.(7) The investigation resulted in the first major lawsuit filed by the department against an institution for violating fair lending laws.(8) This is in sharp contrast to the number of suits filed for civil rights violations in other areas, for example, housing and employment. Congress also responded to repeated claims of lending discrimination by amending the Fair Housing Act in 1988 to allow private parties to originate mortgage discrimination lawsuits more easily. The ECOA was amended in 1991 to require bank regulators to refer cases to the Department of Justice instead of handling them independently, sending a signal that the department was going to be more aggressive in the prosecution of such cases. Perhaps most significantly, the 1975 HMDA was amended in 1988, 1989, and 1991 to develop a database that would provide regulators and the public with data to analyze depository institution lending patterns.

As originally enacted, HMDA required depository institutions and their subsidiaries to provide the total number and dollar value of mortgages originated and purchased in the local market, typically segmented by census tract. The 1989 amendment required lenders to report information at the loan application level regarding race, gender, and income, along with details on the disposition of the application (deny/accept/withdraw, reason for denial, etc.).(9) Banks were required to make the data publicly available. These expanded data have enabled regulators to complement their manual reviews of loan files with systematic statistical analysis.(10) The additional data also allow the public to more closely scrutinize lending patterns of depository institutions.

There have also been recent efforts by bank regulators to help depository institutions comply with fair lending regulation by clarifying the compliance requirements. While the purpose of fair lending laws and regulations is relatively straightforward, there have been problems in implementation, and disagreements have arisen between regulatory agencies and lenders as to interpretations of the law. To provide guidance, a 1994 interagency task force representing the federal depository regulators released guidelines as to what could constitute discriminatory lending practices.(11) Under these fair lending guidelines, a lender may not, because of a prohibited factor:

* Fail to provide information or services or provide different information or services regarding any aspect of the lending process, including credit availability, application procedures, or lending standards;

* Discourage or selectively encourage applicants with respect to inquiries about or applications for credit;

* Refuse to extend credit or use different standards in determining whether to extend credit;

* Vary the terms of credit offered, including the amount, interest rate, duration, or type of loan;

* Use different standards to evaluate collateral;

* Treat a borrower differently in servicing a loan or invoking default remedies;

* Use different standards for pooling or packaging a loan in the secondary market;

* Express, orally or in writing, a preference based on these prohibited factors, or indicate that it will treat applicants differently based on these factors; or

* Discriminate because of the characteristics of a person associated with a credit applicant or the prospective occupants of the area where property to be financed is located.(12)

While blatant discrimination may be obvious to most parties, there are times when sound business practices may result in an unintended discriminatory practice against a protected group. To emphasize to lenders the need to avoid unintended effects in setting underwriting criteria, the interagency task force also listed the forms of discrimination that the courts had previously recognized as illegal. These include: overt discrimination - the lender openly discriminates; disparate treatment - the lender treats applicants differently based on one of the prohibited factors (whether or not it is motivated by prejudice or intent to discriminate); and disparate impact - the lender applies a practice uniformly to all applicants, but the practice has a discriminatory effect and cannot be justified by business necessity.

As a result of the increased scrutiny of lending practices by regulators, there has been a significant increase in the number of ECOA violations referred to the Department of Justice by the regulatory agencies and in the number of suits filed by the department for violation of the fair lending laws. Most of the suits have been settled through well-publicized consent agreements, which relayed the message of stringent enforcement of the fair lending laws. In evaluating the effect of the fair lending laws on mortgage activity, therefore, one would expect to see more of an impact on lending patterns in the 1990s, as institutions respond to increased regulatory pressure.(13)

The CRA

The major impetus for the 1977 passage of the CRA was concern by community groups that banks and thrifts were not responding adequately to the credit needs of local communities. Depository institutions were accused of discriminating against individuals based on the characteristics of their neighborhood, that is, redlining. This was seen as having a particularly adverse impact on minority groups and contributing to the deterioration of inner-city neighborhoods. However, the emphasis of the act was on adequately preserving communities and not on channeling credit based on race. Community groups argued that it was common for banks to reinvest a relatively small portion of deposits generated from local communities back into those markets.(14)

The initial community reinvestment bill was much more intrusive to banks than the final act. The initial proposal argued that banks were chartered institutions with access to a government safety net and, as such, had a formal responsibility to perform social functions in addition to pursuing the objectives of a private enterprise. The proposal defined the bank's relevant local market from which it received deposits and required it to focus on satisfying credit demands in this market prior to exporting funds to other areas. Banks argued that such behavior would run counter to existing safety and soundness regulation and constituted overt credit allocation without regard to the credit quality of applicants in different geographic areas.

The final act omitted the explicit credit allocation criteria. It required financial institutions to serve the convenience and needs of the communities in which they were chartered without mandating how this was to be accomplished. Additionally, it emphasized the need for bank management to be conscious of community credit needs and stressed that this was to be done without sacrificing safety and soundness.

The mandate of the CRA, to have institutions serve the needs of the community in which they are chartered, was actually already in place. The 1935 Banking Act required banks to meet the convenience and needs of their communities, as did the 1956 Bank Holding Company Act and the bank charter itself. The fair lending laws, while not explicitly outlawing redlining, addressed similar concerns. Finally, while HMDA provided no mechanism for imposing sanctions on depository institutions, the data were being collected precisely for the purpose of monitoring lending patterns and detecting neighborhood redlining. The real thrust of the CRA was to reemphasize the need for good lending practices, to shift the emphasis on reinvestment away from the liability side of the balance sheet (deposit gathering) to the asset side (credit generation), and to put the onus squarely upon regulators to monitor the lending patterns of financial institutions and encourage investment in local communities.

In the early years of the CRA, regulatory agencies required banks to specify their local community; develop a public statement, including the local community definition and listing the type of credit instruments the bank intended to provide; post a list of consumer rights under the CRA; and maintain a file of public comments for public inspection. These procedural requirements were relatively straightforward. In addition, regulators performed an evaluation to "assess the institution's record of meeting the credit needs of the entire community, including low- and moderate-income neighborhoods, consistent with the safe and sound operation of each institution" (Regulation BB).

To assess the institution's performance in satisfying this requirement, the regulators developed 12 assessment factors grouped into five performance categories:(15)

Category A: Ascertainment of community credit needs

1. Communication with members of the community to ascertain credit needs; and

2. Extent of involvement by the board of directors in the CRA activities.

Category B: Marketing and types of credit offered and extended

3. Marketing efforts to make the types of credit offered known in the community;

4. The extent of loans originated in the community; and

5. The extent of participation in government loan programs.

Category C: Geographic distribution and record of opening and closing offices

6. The geographic distribution of credit applications, approvals, and denials; and

7. The record of office openings and closings and extent of service provided at the offices.

Category D: Discrimination and other illegal credit practices

8. Practices to discourage credit applications; and

9. Discriminatory or other illegal practices.

Category E: Community development

10. Participation in community development projects or programs;

11. The institution's ability to meet community credit needs; and

12. Other relevant factors which could bear upon the extent to which the institution is helping to meet the credit needs of the community.

For each of the assessment factors, the examiner was to assign a grade of 1 (exceptional) to 5 (significantly inferior), similar to the CAMEL rating given for safety and soundness evaluation. Later, to avoid confusion with safety and soundness ratings, the CRA rating was changed to a four scale grading system: outstanding, satisfactory, needs to improve, or substantial noncompliance.

The regulations did not impose explicit sanctions on institutions found not to have adequately served the needs of their communities. Instead, the regulator was to consider the CRA rating along with other factors, such as safety and soundness, when ruling on an application for a geographic expansion of facilities through a merger or acquisition, the introduction of new branches, an office change, etc. However, there are additional costs from having a poor CRA rating or being accused of poor CRA performance, even if the application is ultimately approved. For example, the application process can be significantly lengthened and complicated if community groups protest the application. In a period in which banks were aggressively expanding geographically, the potential for lost deals, delays in expansion, and negative public relations could be quite burdensome.

Following passage of the act, bankers frequently complained about the vagueness of the requirements, including the lack of a specific ranking or weighting scheme for the assessment factors to guide the allocation of resources. Regulatory agencies would periodically issue policy statements providing guidance to institutions as to how the assessment criteria were scored and discussing elements of effective CRA programs. Most of these statements emphasized effort, and the documentation of such effort, instead of performance. In the late 1980s, Congress amended the act to have the assessments made public and increased public scrutiny of banks and regulators.

As with the fair lending laws, enforcement of the CRA intensified in the early 1990s. Denials of merger or acquisition applications based on poor CRA performance became more common. Although the ratings were not made public prior to 1990, evidence suggests that regulators have tightened enforcement and have been more strict in assigning CRA ratings.(16) To stress the commitment to low-income financing, Congress passed the Federal Housing Enterprises Financial Safety and Soundness Act in 1991. This act put the Federal National Mortgage Association and the Federal Home Loan Mortgage Corporation under an affirmative obligation to facilitate financing of low- and moderate-income housing. It also established mortgage purchasing goals for these agencies relating to low- and moderate-income families for affordable housing and for the central city. Bankers continued to complain about the vagueness of CRA requirements and the resulting regulatory burden. Community groups continued to complain that banks were inadequately serving the credit needs of their local communities and that regulators were inadequately enforcing the act.

TABLE 1

CRA test ratings

Component test ratings are assigned to reflect the bank's lending,
investment, and services.

Component
test ratings Lending Investment Service

Outstanding 12 6 6

High satisfactory 9 4 4

Low satisfactory 6 3 3

Needs to improve 3 1 1

Substantial
noncompliance 0 0 0

Preliminary composite rating is assigned by summing the three
component test ratings and referring to the chart below.

Points Composite assigned rating

20 + Outstanding

11-19 Satisfactory

5-10 Needs to improve

0-4 Substantial noncompliance

Note: Adjustments to preliminary composite rating - no bank may
receive a composite assigned rating of satisfactory or higher
unless it receives at least low satisfactory on the lending test.

After much public and congressional debate, new CRA regulations were issued in 1995 for implementation over the following two years. The new regulations stressed performance over effort in meeting CRA requirements and introduced a new evaluation system, replacing the previous 12 assessment factors with three new tests: lending, investment, and service. For each test a bank is assigned one of five grades from outstanding to substantial noncompliance. There is also an overall composite rating for CRA compliance.(17)

The lending test evaluates whether a bank has a record of meeting the credit needs of its local community. The regulator evaluates the number, amount, and distribution across income groups and geographic areas of mortgage, small business, small farm, and consumer loans in the assessment area(s) or communities.(18) The regulator also considers the innovativeness of the bank in addressing the credit needs of low- or moderate-income individuals or areas and in generating community development loans. As illustrated in table 1, the lending test carries a disproportional weight in determining the composite rating. A bank cannot receive a composite rating of satisfactory or better unless it receives a minimum of low satisfactory on the lending test.(19)

The investment test evaluates how well a bank satisfies the credit needs of its local neighborhoods through qualified community investments that benefit the assessment area. Again, the bank's innovativeness in responding to community development needs is also taken into account.

Finally, the service test evaluates how well the credit needs of the community are met by the bank's retail service delivery systems. This includes the distribution of branches across areas serving low- to moderate-income individuals and geographies, as well as alternative delivery systems, such as ATM, telephone, computer, and mail. The delivery systems and services should be directed at meeting the needs of the local community, for example, low-balance checking accounts and extended lobby hours. Again, the innovativeness of the bank in using these alternative delivery systems to serve the low- and moderate-income individuals and neighborhoods within the community is taken into consideration.

Although it is too early to determine the effectiveness of the revisions to the CRA, a recent Government Accounting Office review of the new guidelines argued that regulators may face significant challenges in implementing the reforms.(20) The potential problems are similar to those which existed before the reforms, namely:

* A continued need for excessive documentation of effort and process;

* Inconsistency in ratings and uncertainty about the performance criteria;

* Incomplete consideration of all relevant material in determining the performance of the institution; and

* Dissatisfaction with regulatory enforcement (depending on one's perspective, too stringent or too lax).

To minimize potential problems, the report recommended that significant efforts be made to provide improved examiner training, improve the quality of the data used in evaluating performance, and increase the use of public disclosure of the ratings.

Evidence of discrimination in mortgage lending

The CRA was introduced because redlining was believed to be a common practice by banks. The fair lending laws were passed because there was a perception that certain borrowing groups were not being treated equitably. However, there continues to be significant disagreement as to the extent of these problems.

Housing and mortgage discrimination has been a topical issue since the 1960s, when community groups argued that neighborhoods were deteriorating as a result of practices by mortgage originators. The originators were accused of using noneconomic criteria to limit funding to non-white applicants and/or nonwhite neighborhoods. Research in this area has intensified in recent years as amendments to HMDA reporting requirements have increased the availability of data used to compare lending patterns across race and ethnic groups, income groups, and geographic areas. However the data exclude many of the more relevant variables used in the credit evaluation process.(21) The most meaningful studies of the role of race and neighborhood effects in mortgage lending incorporate information beyond HMDA data and evaluate discrimination based either on the neighborhood of the applicant or the characteristics of the individual applicant. These studies are divided into four classes: neighborhood redlining studies, application accept/reject studies, studies of default rates, and performance of institutions specializing in loans to low-income individuals or in low-income neighborhoods. Below, we summarize the studies to emphasize the ongoing controversies in this area of research.

Redlining studies

Redlining is the practice of having the loan decision based on, or significantly influenced by, the location of the property without appropriate regard for the qualifications of the applicant or the value of the property. As a result, the neighborhood's financial needs are not adequately served and the region is unable to develop economically. Redlining studies typically take the neighborhood as the unit of observation, evaluating whether the aggregate supply of funds made available is related to the racial composition of the area.

Early analysis of differences in loan originations across markets found significant differences based on the racial composition of the neighborhood. However, these studies attributed all market differences to the race variable.(22) The findings from a number of recent studies, which either directly or indirectly addressed the redlining issue and attempted to explicitly account for market differences, are summarized in box A.

Although improvements have been made in redlining studies, inherent methodological problems remain. First, in a number of redlining studies the unit of observation may be too large. To the extent redlining occurs, it could be for a relatively small area, such as two or three city blocks. In larger areas, such as metropolitan statistical areas (MSA), redlining may not be detectable in the aggregate data. Additionally, assuming some lenders redline and others do not, if borrowers eventually find the non-redlining lender, data at the broader level will imply that no redlining has occurred. The unit of observation should, therefore, be relatively small. There may also be a significant omitted variable bias. Exclusion of variables correlated with race may produce a significant coefficient for race even in the absence of discrimination. A standard criticism of redlining studies is that they inadequately account for demand factors. Thus, it is impossible to attribute differences in mortgage activity across markets to an inadequate supply of funding (redlining) or to a lower demand from potential borrowers. The creditworthiness of the applicant pool is also important since the riskiness of the loan will obviously be a determining factor in the underwriting decision. Additional variables to account for differences in borrower credit demand and creditworthiness that have been included in the recent studies are neighborhood average income, percent of owner-occupied houses, changes in property values, poverty and welfare rates, percent of housing units vacant, crime rates, wealth measures, mobility rates, average age of population and housing stock, total housing units, duration of residency, and the stock of conventional mortgages.

Typically, studies that have accounted for these market characteristics more comprehensively have reported a less significant impact of racial composition than that found in earlier studies. For example, when Holmes and Horvitz (1994) excluded measures of risk in their analysis of the Houston market, they found that the flow of mortgage credit was negatively associated with minority status, consistent with redlining. When the risk measure was included, minority status was not found to influence the flow of credit.(23) Studies which employ a single-equation model to explain the amount of credit made available in a neighborhood will be mixing elements of both supply and demand for credit. Redlining will affect the supply loans. However, with the single-equation approach the supply and demand effects cannot be separated (Yezer, Phillips, and Trost, 1994), Arguing that the race variable represents discrimination requires that there be no demand-side effects. As mentioned above, a number of studies have shown this to be incorrect. Finally, model specification has been shown to drive some results (Home, 1997). Concern with model specification argues that one should use a relatively flexible financial form which has the more commonly used alternative forms nested within it.

Some researchers have argued that the problems associated with the above credit flow type of redlining studies are too large to overcome and, as a result, these studies cannot adequately identify the role of racial composition of the neighborhood in loan decisions. An alternative approach, which addresses the problem of individuals eventually finding the non-redlining lender, is to directly survey individuals who were active in the mortgage market. Benston and Horsky (1992, 1979) surveyed home sellers and buyers to gather information on credit difficulties encountered in attempting to sell or purchase homes in several U.S. cities. Instead of viewing only the mortgages approved, the survey gathered information on individuals who requested credit but were unable to obtain it (for reasons such as redlining), in areas in which charges of redlining had been made and in control areas. If obtaining credit was a problem, additional information as to the reason for the problem was obtained - for example, unemployment, inadequate down payment, or location of the house. The survey explicitly asked home buyers if either a lending institution or real estate agent had stated or implied that obtaining a mortgage might be difficult because of the neighborhood in which the home was located. In both studies, the authors were unable to detect evidence of discrimination or unmet demand. The bottom line appears to be that there is little convincing evidence to suggest that redlining explains lending patterns in low-income neighborhoods.

Accept/Reject studies

Given the above criticisms of credit flow studies, the availability of more detailed HMDA data since 1990, and a desire to more directly address the discrimination issue, recent research has taken a more microeconomic approach. Accept/reject studies take individual application data and evaluate the determinants of the lender's decision. They estimate a probability of rejection function based on various risk factors and include a race variable to account for discrimination.(24) While these studies can also be used to test for redlining, their focus is on discrimination with respect to individual applicants. (For a sample of these studies, see box B.)

Prior to the availability of HMDA data, Black et al. (1978) used special survey data to determine the economic variables important to the lending decision and whether personal variables such as race played a role. After accounting for economic variables and terms of the loans, they found that, although the personal characteristics did not significantly add to the power of their model in explaining the accept/denial decision, race was significant. Black applicants had a higher probability of denial at the 90 percent significance level.

In a well-publicized accept/reject study, Munnell et al. (1992) used HMDA data augmented with survey information about the creditworthiness of borrowers to analyze lending behavior in Boston. A variable to account for the racial composition of the market was not found to affect the lender's accept/reject decision, but applicant race was found to be statistically related to the decision. Minorities were rejected 56 percent more often than equally qualified whites.

The Boston study has been criticized for a number of reasons.(25) First, as with the credit flow studies, there is the potential for omitted variable bias. If omitted variables are associated with the race variable, the coefficient on race will account for the true effect of race plus that of the omitted variable(s). The Boston study included several variables to account for borrower risk. However, not all risk factors could be captured, and some researchers argue that the race coefficient is actually capturing the riskiness of the applicant. Race would appear significant in an analysis which fails to account for wealth if, as has been shown elsewhere, minorities have lower levels of wealth. There was also little consideration of the characteristics of the property and credit history of the applicant. Second, the study has been criticized for data errors. These potential data errors include monthly incomes that are inconsistent with annual levels, negative interest rates, loan to value ratios exceeding one, loan to income ratios outside reasonable ranges, the inclusion of black applicant denials because of over-qualification for special lending programs, and a number of extreme outliers. Brown and Tootell (1995) and Munnell et al. (1996) contend that even after accounting for the data concerns, the fundamental result remains - minorities are more likely to be denied mortgages than similarly qualified whites.

Other follow-up studies have shown mixed results. Using data from Munnell et al. (1992), Zandi (1993) found no race effect, while Carr and Megbolugbe (1993) found the effect remains after "cleaning the data," as did Glennon and Stengel (1994). Using a model similar to that in Munnell et al. to evaluate the Boston and Philadelphia markets, Schill and Wachter (1993) found evidence consistent with redlining and discrimination. When variables are included to proxy for neighborhood risk, the neighborhood racial composition became insignificant, although racial status still significantly decreased the probability of acceptance. Stengel and Glennon (1995) also found that it is important to use bank-specific guidelines in the analysis to capture unique, but economically based, underwriting criteria. Using a more generic market model, for example, secondary market criteria, can lead to misleading results.(26) Using cleansed data from Munnell et al. (1992), Hunter and Walker (1996) did not find evidence of discrimination via higher underwriting standards for all minorities. They contend that race matters only in the case of marginally qualified applications. Needless to say, there is little uniformity of view.

Yezer (1995) and Rosenblatt (1997) argue that fundamental problems in the use of accept/reject models to evaluate discrimination result from the informal prescreening of applicants. Both applicants and lenders only want to proceed with applications that appear likely to qualify for a loan because denials are costly for both parties. Thus, during the initial lender-borrower contact, the lender and borrower decide whether the application warrants pursuing. Then, the formal application takes place, and denials occur only in those cases in which information not available in the initial contact affects the decision (for example, bad credit history). Therefore, denial may be as closely related to communication skills and cultural background as to economic variables. Sophisticated underqualified potential applicants will not reach the formal application process because they realize they will not be accepted, while unsophisticated candidates will follow through only to be denied. Thus, there is a significant selection bias problem in the formal application stage which may explain the race differentials. To support this view, Rosenblatt (1997) cites evidence that education levels are strongly predictive of credit approvals. The argument, therefore, is that the information in denial rate data may not be what researchers perceive it to be.

Default rate studies

An alternative means of evaluating lender discrimination is to examine the default rates of borrowers thought to be discriminated against relative to other borrowers. Researchers have compared default rates across groups based on the theory that if minorities are overtly discriminated against, the average minority borrower should be of higher credit quality than the average nonminority borrower.(27) This should be reflected in mortgage default rates and resulting loss rates; for minority loans, both should be lower. However, studies have not found evidence of lower default rates for minority holders of mortgages. In critiquing the Boston study, Becker (1993) cited data indicating the default rates were equal for white and minority sections of the Boston market, which was not consistent with overt discrimination.(28) A more recent study by Berkovec et al. (1996) also tests for discrimination using default rates. Controlling for various loan, borrower, and property related characteristics, the authors evaluated the default rates and resulting losses for FHA-insured loans and found a higher likelihood of default on the part of black borrowers and higher loss rates. These results suggest that lenders, perhaps as a result of regulatory pressure, may have over-extended credit to minorities.

However, this line of research has also been criticized. First, if discrimination occurs, while the marginal minority borrower may be better qualified than the marginal white applicant, inferences about the average borrower cannot be made without making assumptions about the distribution of creditworthiness across the two groups of potential borrowers, for example, Ferguson and Peters (1995). The distributions could be significantly different. Additionally, minorities may also be treated differently once they are in default. Default studies typically use data on foreclosures. Bank forbearance in defaults favoring one of the two groups could bias the results.(29)

Performance studies

There are two general areas of research relating bank performance to the CRA and fair lending regulations. The first deals with the profitability associated with lending in low-income markets. If such lending is not profitable, regulation requiring it should adversely affect performance. The second area of research addresses the implications of mortgage discrimination on bank performance. If some banks are choosing to discriminate and forego profitable lending opportunities, other banks that do not discriminate should be able to exploit these opportunities.

During the debate prior to the enactment of the CRA, critics argued that economics was driving lending patterns and the CRA might either have no impact, but be costly to implement, or actually generate bad loans. From the banks' perspective it would be a tax and, if lending patterns did not change, it would be without benefits. If increased lending in the low-income market did occur, but was not as profitable as that in alternative markets, then the CRA would act as a tax and a credit redistribution mechanism. The argument in favor of the CRA was that banks were foregoing profitable opportunities because of discriminatory behavior or market failure, and performance could be enhanced if they became more actively involved in this market (although performance could be adversely affected in the short run as start-up costs were incurred).

There have been a limited number of studies evaluating the effect on performance of lending in the low-income market. Canner and Passmore (1996) offered a number of testable hypotheses concerning the potential impact on profitability and the relationship between the extent of the bank's activity in this market and performance. They found no evidence of lower profitability at banks specializing in the low-income market, consistent with the view that, once start-up cost are incurred, lending in this market can be just as profitable as in other markets.(30) Beshouri and Glennon (1996) evaluated the relative performance of credit unions that specialize in the low-income market and found that while these specialized firms have greater return volatility, higher delinquency rates, charge-off rates, and operating costs, they are compensated for these differences and generate similar rates of return. Similarly, in analyzing the performance of low-income and minority lending, Malmquist, Phillips-Patrick, and Rossi (1997) found that while low-income lending was more costly, lenders were compensated with higher revenues, making profits similar for both low- and high-income lending. Finally, Esty (1995) evaluated the performance of Chicago's South Shore Bank, which has been held up as the model community development bank with the dual objectives of making a profit and aiding in the development of the local community. Esty's analysis found the economic return of the bank to be substandard. Shareholders, however, appeared to be willing to trade off the lower return for the social return received from community improvement. That is, the shareholders' objectives were apparently aligned with the dual objectives of the bank.(31) In interviews with shareholders and employees, Lash and Mote (1994) found similar evidence of a willingness to trade off economic profit to emphasize the development objective. While the behavior of South Shore's management and shareholders may be admirable, if Esty's analysis is correct, it is not obvious that this model can be implemented across the entire industry.

The second performance-related area of research deals with the profit implications of discrimination. If an institution overtly discriminates, it will deliberately forego profitable lending opportunities. This implies that lenders that do not discriminate will be the beneficiaries of this behavior. Assuming that minority-owned banks do not discriminate against minorities, one might expect them to outperform the discriminating-banks. Calomiris, Kahn, and Longhofer (1994) developed a model of cultural affinity to explain differences in minority denial rates. Their basic argument is that because of a general lack of familiarity with the culture of minority applicants, the typical white loan officer may not be as accommodating with these applicants as he would with a white applicant. For the minority applicant, the loan officer will rely more heavily on low-cost, objective information instead of making the extra effort, as with the white applicant, to obtain additional information to improve the chances of approval. There is some empirical support for this argument (see Hunter and Walker, 1996). Again, this implies that minority-owned banks should benefit, since they will not lack a cultural affinity with minority applicants.(32)

If discriminatory banks forego profitable opportunities, ceterus paribus, minority-owned banks should have superior profitability, lower minority denial rates, and lower bad loans. However, the empirical evidence does not support this. A number of studies have found that minority-owned banks have lower profits (Bates and Bradford, 1980, Boorman and Kwast, 1974, and Brimmet, 1971). There is also evidence of higher loan losses at minority-owned banks (Kwast, 1981). Additionally, there is evidence that bank ownership shifts from white to black control result in fewer loans being generated (Dahl, 1996). Generally, there is evidence that minority-owned banks do not have particularly good performance or lending records and have relatively poor CRA ratings (Kwast and Black, 1983, Clair, 1988, and Black, Collins, and Cyree, 1997). This evidence is not consistent with overt discrimination.

In summary, the findings for the various forms of discrimination are quite mixed. While some studies have found race to be a factor in loan decisions, the evidence is far from conclusive. Additionally, methodological problems bring into question the validity of many studies. Parties on either side of the issue frequently draw uncritically on the studies that align with their own position. Additional research is needed before we can draw meaningful conclusions about race and the credit decision.

Recent trends in mortgage activity: The effect of regulation

How successful has the recent enforcement of the CRA and the fair lending laws been? Headlines proclaiming surges in credit to minority groups suggest that the stricter enforcement of the CRA and fair lending laws during the 1990s has been successful. Most of these claims are based on recent trends in lending to low-income individuals or in low-income neighborhoods, such as those presented in figure 1.(33) Between 1990 and 1995 the annual number of mortgage originations to low- and moderate-income households, in low- and moderate-income census tracts, and to minorities almost doubled. New loans in census tracts where minorities accounted for at least half the population also grew significantly. As figure 1 shows, there was a considerable increase in the number of loans to individuals targeted by fair lending and CRA regulations. Some bankers argue that the regulatory mandate to increase lending in low-income neighborhoods and to low-income individuals has actually been a blessing in disguise as it has opened up new, lucrative, previously untapped markets. Others continue to criticize the regulations.(34)

A full assessment of the success of the CRA and the fair lending programs would require a comprehensive cost-benefit analysis. Accurately quantifying the cost is difficult.(35) It is also difficult to quantify the success of these programs because of the vagueness of the legislation and the regulations enforcing it. The mandate to banks was to use the proper criteria in making loan decisions and to reach out to the local community, including low- and moderate-income neighborhoods and individuals. Based on this mandate, success may not require any change in lending patterns. Another problem with associating recent lending patterns with regulation is the lack of a control group. The issue is not whether lending to the targeted groups increased, but whether it increased as a result of the regulations,

There are, however, a number of credit flow measures often associated with the CRA and fair lending laws. Concerning redlining, one would want to analyze changes in the volume and dollar value of loans flowing into low-income or minority neighborhoods. The number of applications in these areas could also be considered if redlining resulted in applications not being accepted from these areas. Some argue that the purpose of the regulations was to increase the flow of credit to specific groups of borrowers (based either on income or race), therefore, credit flows to those particular groups could be analyzed. It has also been argued that the elimination of differences in denial rates may be desirable.(36) Concerning fair lending, one would want to analyze changes in lending to minority individuals and/or denial rates.

Although data limitations hamper the degree to which rigorous analysis of the regulatory impact can be undertaken, we can evaluate lending patterns and check for trends consistent with what are typically perceived to be desired changes. We use three different control group specifications. First, we compare lending patterns pre- and post- the recent regulatory changes. Second, we compare the degree of lending to targeted groups (minority and low-income individuals and neighborhoods) with lending to nontargeted groups. Last, we compare the lending behavior of more heavily regulated depository institutions with that of less regulated mortgage companies.

Below, we present evidence on these credit flows and discuss how they align with the goals of the legislation. We would expect the CRA and fair lending reforms of the late 1980s and early 1990s to have increased lending to minority and low-income individuals and low-income neighborhoods. We would also expect that most of the impact would be concentrated in recent years as regulatory, legal, and public scrutiny intensified and the cost of failing to satisfy the requirements increased. We analyze two potential effects. If the regulations were successful in getting lenders to expand their business into new markets, this might influence the overall level of mortgage activity. Alternatively, we may see distributional effects as lenders allocated a larger share of the credit pool toward the new markets.

To analyze the effect on the aggregate level of mortgage activity, we used a nominal dollar measure of all mortgage activity for 1970 through 1995, combining originations and refinances, from the Survey of Mortgage Lending Activity issued by the U.S. Department of Housing and Urban Development (HUD) [ILLUSTRATION FOR FIGURE 2 OMITTED]. Figure 2 suggests considerable growth in mortgage originations over the 1990s, in particular 1993-94. Can the high rate of growth in the 1990s be attributed to regulatory-induced changes in lending behavior or to other factors related to the aggregate demand for housing credit? Three pieces of evidence suggest that the latter hypothesis may be more accurate. First, there is considerable growth throughout the entire period, even in real terms as indicated by the colored line depicting the dollar value of mortgage originations deflated by the Consumer Price Index for Urban Workers. The colored line also highlights the procyclical and seasonal nature of mortgage originations. Second, there is a substantial decline in the number of mortgage originations after the 1993 peak, which might be due to a curtailment of refinance activity.(37) Clearly, there is no regulatory explanation for this decline. There are probably a number of factors beyond regulation that affect the number of originations and increase the difficulty of graphically detecting a structural break. To address this, we use a regression model of the quarterly growth rate of originations, controlling for the growth rate of gross domestic product (GDP), the change in mortgage rates, and the growth rate of the consumer price index.(38) Quarterly indicators are included to absorb the seasonality in the dependent variable. Table 2 displays the regression results. In column 1, over 50 percent of the variation in the growth of originations is explained by the right-hand-side variables. The coefficients suggest that an increase in [TABULAR DATA FOR TABLE 2 OMITTED] mortgage rates corresponds to a decrease in the growth of mortgage originations, while an increase in the growth rate of GDP corresponds to an increase in the level of originations. The controls for seasonality, the quarterly indicator variables, suggest faster growth in mortgage originations in the second and third quarters than in the first and fourth. Column 2 of the table presents the regression model with the addition of a binary variable to capture a structural shift in the post-1990 period. The coefficient of the post-1990 indicator variable is actually negative, but is significant at only the 74 percent confidence level. Therefore, the regression model is unable to support the hypothesis that mortgage originations were stronger in the period following the recent regulatory changes. Tests for structural breaks for post-1992 and post-1993, columns 3 and 4, produced similar results.

Although we find the recent growth in mortgage lending is consistent with earlier patterns, the regulations may have resulted in distributional changes in lending patterns.(39) That is, there may be a shift in lending emphasis away from traditional markets toward low- or moderate-income groups and individuals. To evaluate this, we assembled HMDA data for depository institutions and their affiliates over the 1982-95 period and decomposed the data by income groups and, when possible, racial groups.(40)

To the extent that regulations influence lending patterns, we have argued above that the effect would be most evident in recent years, because of increased scrutiny, and most pronounced among low-income and minority borrowers and neighborhoods. We therefore divided total lending activity into four income categories and evaluated growth trends in the number of mortgage applicants, originations, and dollar value of originations for the 1990s. We also developed data for the number and dollar value of originations for the 1980s by neighborhood income levels.(41) Table 3 shows the 1990 levels and relative shares of mortgage activity on which our analysis is based. The targeted groups received a relatively small portion of the applications, originations, and dollars. Low- and moderate-income tracts account for approximately 10 percent of the applications and loans and slightly less of the dollars lent. Low- and moderate-income individuals, as opposed to home purchases in low- and moderate-income tracts, represent nearly 20 percent of applications and originations, but still less than 10 percent of the dollars lent. Roughly 80 percent of mortgage activity (applicants, originations, and dollars) involved white applicants. These figures demonstrate that the targeted populations are a relatively small share of the aggregate, which may explain why increases may not be observable in the aggregate data.

TABLE 3

Base mortgage lending data, 1990

 Applications Acceptances Dollars
 (000s) (000s) (billion)

Total Applications 1,491.35 1,276.16 127.72

Tract income/MSA
income shares

Low income 0.01 0.01 0.01
Moderate income 0.10 0.09 0.08
Medium income 0.57 0.57 0.49
High income 0.31 0.33 0.43

Applicant income/MSA
income shares

Low income 0.05 0.04 0.01
Moderate income 0.17 0.16 0.08
Medium income 0.27 0.28 0.20
High income 0.50 0.53 0.71

Race shares

White 0.82 0.84 0.82
Black 0.06 0.05 0.04
Other minority 0.11 0.11 0.13

Source: HMDA data, see footnotes 33 and 40.

The year-over-year growth rates for the number of loans originated by neighborhood income groups for the 1980s and 1990s are presented in figures 3 and 4, respectively. For the 1980s, growth in the low- and moderate-income groups lagged that in other areas. For four of the years in the 1980s, the low-income tracts showed the slowest growth and the moderate-income tracts also showed relatively slow growth. In these years, growth in the overall market was quite robust. Thus, for the 1980s overall, growth in loan volume was not particularly concentrated in the low- and moderate-income groups. Originations to the low- and moderate-income groups grew more than 40 percent between 1985 and 1986, exceeding growth for these groups for any single year throughout the 1990s. However, the 1986 growth of mortgage originations in middle- and upper-income tracts still exceeded that of the low- and moderate-income groups.

Things changed in the 1990s. After 1991, growth was relatively fast in the two lowest income groups, particularly in the years when overall market growth was greatest. This finding suggests that banks have responded to the CRA and have made significantly more loans in the low- and moderate-income markets. The change is overwhelmingly statistically significant based on a test of whether the share of loans in each income category is constant throughout the 1990s. We conclude that the growth in mortgage originations has not been uniform throughout the 1990s, consistent with the 1992 through 1994 growth spurt in lending to low- and moderate-income groups.

Figure 5 presents data for recent mortgage applications.(42) After 1991, low- and moderate-income neighborhoods saw significantly stronger growth than occurred in other areas. Growth in the middle-income group, where the majority of mortgage activity is occurring (see table 3), saw the slowest increase over this period. These trends are consistent with the view that banks have been making a significant effort to encourage applications from lower-income neighborhoods and with statements by community groups that progress is being made in less affluent neighborhoods. The test of differences across categories is again highly statistically significant.(43)

Figures 6 and 7 analyze mortgage activity by the income level of the borrower instead of the neighborhood in which the property is located. While not quite as pronounced as the neighborhood data, figure 6 shows growth in mortgage activity to low- and moderate-income individuals, particularly for the years in which overall growth was greatest. The mortgage application data in figure 7 tell a similar story. Overall, the data suggest an increase in the growth of loan applications and loans approved for low- and moderate-income individuals, with much of the growth coming after 1992. However, the differences are only significant at the 46 percent and 54 percent confidence levels, respectively, for originations and applications. Thus, based on this test, lending to low- and moderate-income individuals was uniform throughout the 1990s. Data on applicant income were not collected for the 1980s so we cannot compare the two periods.

Figures 8 and 9 show loan activity by applicant race. While neither the CRA nor the fair lending laws explicitly require lenders to change underwriting criteria and affirmatively pursue additional minority mortgage business, lenders may believe that doing so will help them avoid charges of discrimination and be looked upon more favorably during their regulatory assessment. The growth in minority applications and originations during the 1990s has been high relative to that for nonminorities. The increase in applications and originations among blacks is even more significant.(44) Figures 10 and 11 present a similar analysis based on the minority proportion of the census tract as opposed to the minority status of the applicant. Since 1991, growth appears to have been relatively similar across the groups.

Several of the results in figures 3-11 are consistent with efforts by banks to target low- and moderate-income individuals and neighborhoods in their mortgage business. This observation is based not on the level of loans made but on the fact that the growth in lending to the targeted groups exceeded that to the nontargeted groups. One could argue, however, that the improvements are somewhat diminished, being from such a small base (see table 3). Lending in low- and moderate-income neighborhoods constitutes approximately 10 percent of total originations and even less of the dollar value of loans originated. Based on income alone, we would expect the demand for mortgages in these neighborhoods to be relatively low. Mortgage activity among low-income individuals constitutes approximately 20 percent of the market. However, the 31 percent growth in mortgage originations in low- and moderate-income tracts from 1993 to 1994 corresponds to nearly 35,000 loans and approximately $2.7 billion. If all of this change is attributed to the regulations, it translates to just over 100 loans and $8 million per MSA.

In addition to the number of loans and applications, we evaluated denial rates for ethnic groups. Minorities have typically been shown to have higher denial rates than other applicants. One of the common debates in the literature and popular press is whether the differences across racial groups can be explained by economic characteristics.(45) We present two measures of differences in denial rates. The first is a standard odds ratio: Based on the loan decision, we calculate the odds of a minority applicant being denied a loan relative to the odds of a nonminority applicant being denied.(46) An odds ratio of 1 corresponds to equality of the denial odds for white and minority applicants; values above I correspond to more minority denials. If minority status is associated with lower creditworthiness, we would expect the odds ratio to be higher because of the differences in qualifications. To partially account for this, we also calculated an odds ratio conditioned on income and loan value.(47) The odds ratios are presented in table 4.(48)

Interpretation of the odds ratios as evidence of discrimination is difficult due to the small number of variables collected in the HMDA data. Instead, we focus on the changes in the ratio over time. Under the assumption of constant quality of the applicant pools, changes in the ratios may be attributed to changes in lender behavior. Cyclical economic changes, however, are likely to affect the creditworthiness of the applicant pool. To account for this, we also calculated the two odds ratio measures for a sample of independent mortgage companies. These are typically thought to be less stringently regulated with respect to the CRA, but they still report for HMDA purposes through HUD. Thus, we use these as a control group that we can contrast with the regulated banks to distinguish the effect of regulation.

As table 4 indicates, for depository institutions the unconditioned odds ratio is relatively stable during the 1990s.(49) The odds of a minority applicant being denied a mortgage request are approximately twice those of a nonminority applicant. The odds ratio and trend estimates are statistically different from 0 at the 99 percent confidence level. Differences between years are typically statistically significant at conventional levels. The conditioned ratio for banks is somewhat similar to the unconditioned measure, but declines throughout the period. The additional information embedded in the conditioned measure does explain part of the difference between the two borrowing groups; however, it leaves much unexplained. Additional information on the creditworthiness of the applicant and any discriminatory effects would be needed to explain the remainder. The declining trend in the odds ratio, after conditioning on a flexible specification of loan amount and applicant income, suggests a change in the treatment of minority applicants relative to nonminority applicants over the 1990-95 period. Essentially, lenders became more accommodating to minorities. The effect is more apparent in the last two columns of the table, where we repeat the analysis using black/white odds ratios. These results are consistent with more stringent regulations producing a change in lender behavior, assuming that the unobserved characteristics of the applicant pool remain constant over time.

The odds ratios for the independent mortgage companies are also presented in table 4. The ratios for these less regulated companies are much more erratic, but display a similar downward trend. Disparities between minority and nonminority applicants and between black and white applicants decline over time for both sets of institutions, suggesting that the change may not be the result of the regulations.(50)

Another way to assess the extent to which the supply of mortgage credit to minorities increased in the mid 1990s is to examine home ownership rates over time. The relaxation of binding credit constraints should cause minority originations and home ownership rates to increase. Figure 12 displays home ownership rates from 1970 to 1994 for white, black, and other minority households. Recent home ownership rates are still well below the rates observed in the early to mid-1980s. Recall that blacks experienced the strongest growth in mortgage originations in 1993 and 1994, yet there was little effect on home ownership. Within our sample, mortgage originations to blacks increased by 23,000 loans from 1992 to 1993 and another 4,000 from 1993 to 1994. In 1994 there were roughly 11.3 million black households in the U.S., implying that 113,000 new loans would be necessary to move the home ownership rate a single percentage point. Viewed this way, the 1993-94 changes appear small.

Summary and conclusions

In this article, we have provided background on the evolution of the CRA and fair lending regulations, summarized the economic literature which pertains to this type of regulation, and presented evidence on the effectiveness [TABULAR DATA FOR TABLE 4 OMITTED] of these regulations by analyzing recent trends in mortgage lending activity. The literature review indicated mixed results. Although early studies appear to find evidence of redlining, more recent studies do not support this finding. Concerning disparate treatment based on the race of the applicant, some studies have found differences in the probability of a loan application being denied that are not explained by economic characteristics. However, these studies have been criticized as having methodological and data problems.

Our major purpose, however, is to provide the reader with a review of the literature and not to take a position on the merits of the various positions concerning whether there is a need for the CRA and fair lending regulations. Regardless of one's position, the regulations do exist and are being enforced. We evaluated how the regulations may alter lending behavior, using a variety of measures of changes in lending patterns. We found that over the 1990-95 period (particularly 1993-94), loan applications and originations increased significantly to groups targeted by the CRA and fair lending legislation. Additionally, it appears there may have been a compositional shift toward blacks and a minor shift toward low-income groups. These changes appear large in terms of growth rates, but they started from a very small base. The changes appear smaller when measured in dollars rather than the number of loans or applications.

It is difficult to attribute the increase solely to the strengthening of the regulations. We assessed the regulatory impact in three ways. First, we used historical trends as a control group. After accounting for economic conditions, we found that aggregate lending in recent years has not been unusually strong. Second, we considered changes in the composition of mortgage activity by examining year-to-year growth rates of applications and mortgage originations. We used the nontargeted groups as controls. The analyses presented some evidence of a change toward increased lending to minority and low-income individuals and neighborhoods. Last, we compared recent trends in denial disparity measures, black/white and minority/nonminority odds ratios over time between depository institutions and mortgage companies. Typically thought to be less stringently regulated, mortgage companies might depict the market result in the absence of additional regulation. Our analysis shows a decline in the odds ratio for both depository and non-depository institutions, suggesting that the effect may not be the result of increased regulatory scrutiny.

Overall, our results are mixed. There is some evidence of changes in lending patterns, and some of the evidence is consistent with changes related to the new regulatory environment. However, the growth rates are not unprecedented and, if entirely attributed to the regulations, translate to approximately 100 loans and $8 million per MSA. We would emphasize that we have not addressed the cost of implementing the regulations relative to the benefits. In addition, we have not addressed the question of whether these credit market regulations are the most effective way of changing the fundamental economic status of the targeted groups. As the regulation of credit markets continues to evolve, it remains important to continually revisit these issues.

NOTES

1 Throughout we use the term bank in a generic sense to encompass all depository institutions.

2 See, for example, Coplan (1996), Lindsey (1996), Seiberg (1996a, 1996b), or Wilke (1996).

3 For a discussion of potential information externalities, see Nakamura (1993) and Calem (1996).

4 Gender was added as a protected category through 1974 amendments, and handicap and family status were added by amendment in 1988.

5 A number of these protected categories were added by amendments in 1976.

6 See Hula (1991) and GAO (1996).

7 See Dedman (1988).

8 For a discussion and examples of increased scrutiny by regulators in more recent years, see Garwood and Smith (1993) and Macey and Miller (1993).

9 The reporting requirement now extends beyond depository institutions to, generally, all lending institutions with assets of more than $10 million with an office in a metropolitan statistical area (MSA) (for depositories) or loan activity in MSAs (for nondepositories).

10 For a description of the process by which the Federal Reserve Banks use statistical models to test for disparate treatment of applicants, see Bauer and Cromwell (1994) and Avery, Beeson, and Calem (1997 forthcoming). For discussion of the use of statistics in detecting mortgage discrimination, see Yezer (1995).

11 See Interagency Regulatory Task Force (1994). Represented in the interagency group were the Department of Housing and urban Development, Department of Justice, Comptroller of the Currency, Office of Thrift Supervision, Federal Reserve System, Federal Deposit Insurance Corporation, Federal Housing Finance Board, Federal Trade Commission, and National Credit Union Administration.

12 Again, although the fair lending laws and the CRA are different, there is significant overlap as evidenced in the last of these items. This is very close to an anti-redlining statement - the very reason for the passage of the CRA.

13 For examples of enforcement activities by the Department of Justice and bank regulatory agencies, see Macey and Miller (1993) and GAO (1996). For a discussion of alternative strategic responses by banks to the increased regulatory pressure from fair lending and CRA regulations, see Evanoff and Segal (1997 forthcoming).

14 This argument was also being made at the time to contest the liberalization of bank branching laws. The concern was that banks would export deposits through their branch network to other, more lucrative, markets.

15 Regulation BB of the Code of Federal Regulations describes the requirements of the community reinvestment regulation.

16 See Macey and Miller (1993) and U.S. Congress (1989).

17 There are alternative tests for wholesale or limited purpose banks, still another streamlined test for small banks, and another for banks choosing to develop and be held accountable for a strategic plan which details how the banks intend to satisfy CRA requirements.

18 Consumer loans will be considered if the bank collects data on this activity and requests that they be considered in the evaluation, or if regulators determine that this activity constitutes a substantial portion of the institution's business.

19 The explicit weight assigned for each test and grade addresses a criticism of the earlier rating system under which bankers frequently complained about the uncertainty as to how they should allocate resources to improve their CRA rating. Emphasizing the lending test also addresses criticisms of the earlier grading scheme by being less process-oriented and more results-oriented.

20 See GAO (1995).

21 For example, HMDA data does not contain information on credit history, wealth, and employment stability of the applicant, or the value or purchase price of the property. Adverse credit history is the most common reason given in the HMDA data for denying loans. For discussions of lending trends in HMDA data, see Canner and Passmore (1994, 1995a, 1995b).

22 A review of much of the early literature can be found in Benston (1981) and Canner (1982).

23 Canner, Gabriel, and Woolley (1991) reported similar findings after controlling for market characteristics. The inclusion of variables to capture risk factors, however, may not resolve the endogeneity problem since their values may be supply induced.

24 Some studies have addressed whether minority groups are "discriminated" against in that they are more likely to receive a particular type of loan which may have less favorable terms. For example, Canner, Gabriel, and Woolley (1991) found minorities are more likely to obtain an FHA loan than are nonminorities.

25 See Horne (1994, 1997 forthcoming), Liebowitz (1993), and Day and Liebowitz (1996).

26 The authors find considerable differences in underwriting standards across banks concerning threshold values for debt ratios, loan-to-value ratios, etc. The authors realize that unlike the more generic market model, regressions with an emphasis on bank-specific underwriting standards will not allow for a direct test of disparate impact.

27 Becker (1971) is typically cited as the source of this argument.

28 Although Becker actually used data from the Boston study, it is questionable if the data are appropriate for critiquing the Boston study based on default rates. The data were for loans originated prior to the period discussed in the Boston study and the analysis was a comparison of default rates across geographic areas based on racial composition. From the data, one cannot conclude how racial default rates compare.

29 For a discussion of these concerns see Tootell (1993) and Ross (1996).

30 The authors, however, emphasize the limitations of their analysis. For example, information on the profitability of low-income lending is not available. Thus, the authors are comparing overall profit levels across banks with different levels of low-income lending although this lending is typically a relatively small portion of the overall portfolio. There have also been concerns expressed recently about the growing number of special lending programs to accommodate the targeted groups and the resulting high default rates, see Seiberg (1996b).

31 Esty also evaluated the impact South Shore had on the local community and argued that there was no evidence of any unique positive relative impact on the local community.

32 One could make the argument that banks with a significant number of minority loan officers could also benefit from a cultural affinity. We do not have data to directly test for this.

33 Low-income neighborhoods are defined as census tracts where the median income is less than 50 percent of the MSA median income. The moderate-income category corresponds to greater than or equal to 50 and less than 80 percent, the middle-income category corresponds to greater than or equal to 80 percent and less than 120 percent, and the upper-income category corresponds to at least 120 percent of the MSA median income. Similar break points define the categories for applicant income relative to MSA income. Tracts are also classified by minority composition into low (less than 25 percent minority), moderate (25 percent to 50 percent), middle (50 percent to 75 percent) and high (75 percent to 100 percent).

34 See, for example, Wilke (1996), Seiberg (1996), or Lindsey (1996).

35 Barefoot et al. (1993) attempt to quantify the compliance cost associated with consumer protection regulations including the CRA. They note that bankers find the CRA to be one of the most onerous regulations faced by banks.

36 A narrowing of denial rate differentials is frequently cited in the popular press as a measure of progress in fair lending, for example, Coplan (1996). Lawmakers have also argued that although some differential may be warranted based on credit quality, the current differences are too great and imply some discrimination is being undertaken: for example, in U.S. Congress (1989), see Illinois Senator Dixon's opening comments during hearings on the CRA.

37 Refinancing and new originations cannot be separated in the HUD data.

38 Changes in mortgage rates measure the contract rate on 30-year fixed rate conventional mortgage commitments reported by the Federal Home Loan Mortgage Corporation.

39 For critiques of the CRA as a means of accomplishing this redistribution, see Lacker (1995), Macey and Miller (1993), and White (1993).

40 It is important to emphasize that our sample may differ from HMDA data reported elsewhere because, unless noted, we are viewing lending activity only for depository institutions or their affiliates, and we are screening out observations that cannot be classified into income and racial subgroups which we expect to be affected by the regulation. By analyzing this group of institutions, we have a homogenous group through time. Certain mortgage originators were only added to the HMDA in recent years. Our sample consists of loan applications for one-to-four family, owner-occupied properties where the mortgage is valued between $1,000 and $1 million, the applicant's income is less than $1 million, and the loan-to-income ratio is less than five. For the 1980s these requirements were imposed on market (census tract) averages since individual loan data were not available on HMDA until 1990. We only consider mortgages for properties in MSAs, and we require complete data on the location of the property - state, county, MSA, and census tract - to allow us to combine HMDA data with census information concerning neighborhood income and composition. Reporting institutions must report on applications for property located in an MSA where they have an office. If the property is outside of an MSA or in an MSA in which the institution does not have an office, it has the option of reporting the MSA information. Thus, some loans made in MSAs will be omitted from our sample because the bank chose not to include the information. The MSA information is necessary to merge the data with census information. Finally, the data must pass the validity checks (Board of Governors 1995).

41 The data are not precisely comparable to that for the 1990s because mortgage originations for purchase and refinancing were not separated during this earlier period. However, the role played by "refis" in the 1980s is not expected to be nearly as erratic as in the 1990s. All analysis of the 1990s excludes refinances.

42 Application data are not available for the 1980s.

43 We reject the hypothesis that shares by income categories are constant over the 1990-95 period based on a chi-square test statistic of 53 with 15 degrees of freedom.

44 The increase in mortgage activity for blacks was also spread across all income levels. The 1990-95 growth rates for blacks in low-, moderate-, middle-, and upper-income neighborhoods were 157 percent, 129 percent, 99 percent, and 101 percent, respectively. In their evaluation of redlining Holmes and Horvitz (1994) found that, after accounting for neighborhood characteristics including risk, more credit was being made available in certain minority neighborhoods. They found a systematic preference on the part of insured lenders (the FHA) toward lending in areas of high or growing minority populations. They attribute this to the pressures created by the CRA and community groups to increase lending in these areas. Similarly, Malmquist, Phillips-Patrick, and Rossi (1997 forthcoming) found that while low-income lending was more expensive, lenders were also compensated with higher revenues making profits similar for both low- and higher-income loans. However, the authors found that profits were inversely related to the share of mortgages originated to blacks. They suggest this is a result of firms "bending over backwards" to yield to regulatory pressures to make more minority loans.

45 Minorities are defined as non-whites. Asians are an exception and typically do not have high denial rates.

46 Odds are defined as the ratio of the probability of denial to the probability of acceptance.

47 We calculate this based on a logit regression which includes, in a flexible functional form, applicant income, loan size, and race.

48 The denial rates may be related to other factors, for example, geographic differences. In our analysis, we emphasize changes through time and assume the effects from other factors are constant across time.

49 The mortgage company data should be interpreted cautiously, since these institutions do not go through the same rigorous editing process and resubmission of revised data as the depository institutions.

50 Alternatively, these "nonregulated" firms can be prosecuted by the U.S. Department of Justice for fair lending violations, so they may not be immune to this regulation.

REFERENCES

Ahlbrandt, Jr., Robert S., "Exploratory research on the redlining phenomenon," AREUEA Journal, Vol. 5, 1977, pp. 473-481.

Art, Robert C., "Social responsibility in bank credit decisions: The Community Reinvestment Act one decade later," Pacific Law Journal, Vol. 18, July 1987, pp. 1071-1139.

Avery, Robert B., Patricia E. Beeson, and Paul S. Calem, "Using HMDA data as a regulatory screen for fair lending compliance," Journal of Financial Services Research, Vol. 11, 1997 forthcoming.

Avery, Robert B., Patricia E. Beeson, and Mark S. Sniderman, "Underserved mortgage markets: Evidence from HMDA data," Federal Reserve Bank of Cleveland, working paper, No. 94-16, 1994.

Avery, Robert B., and Thomas M. Buynak, "Mortgage redlining: Some new evidence," Economic Review, Federal Reserve Bank of Cleveland, Summer 1981, pp. 18-32.

Barefoot, Marrianan & Associates, Inc., A.V. Thakor, and J. C. Beltz, "Common ground: Increasing consumer benefits and reducing regulatory costs in banking," Herbert V. Prochnow Educational Foundation, monograph, 1993.

Bates, T., and W, Bradford, "An analysis of the portfolio behavior of black-owned banks" Journal of Finance, Vol. 35, 1980, pp. 753-768.

Bauer, Paul W., and Brian A. Cromwell, "A Monte Carlo examination of bias tests in mortgage lending," Economic Review, Federal Reserve Bank of Cleveland, Quarter 3 1994, pp. 27-44.

Becker, Gary S., The Economics of Discrimination, Chicago: University of Chicago Press, 1971.

-----, "The evidence against banks doesn't prove bias," Business Week, April 19, 1993.

Benston, George J., "Mortgage redlining research," Journal of Bank Research, Vol. 12, Spring 1981, pp. 8-23.

Benston, George J., and Dan Horsky, "Redlining and the demand for mortgages in the central city and suburbs," Journal of Bank Research, Vol. 10, Summer 1979, pp. 72-87.

-----, "The relationship between the demand and supply of home financing and neighborhood characteristics: An empirical study of mortgage redlining," Journal of Financial Services Research, Vol. 5, 1992, pp. 235-260.

Berkovec, James A., Glenn B. Canner, Stuart A. Gabriel, and Timothy H. Hannan, "Mortgage discrimination and FHA loan performance," Proceedings of a Conference on Bank Structure and Competition, Chicago: Federal Reserve Bank of Chicago, May 1995, pp. 218-237.

-----, "Mortgage discrimination and FHA loan performance," Cityscape, Vol. 2, February 1996, pp. 9-24.

Beshouri, Christopher P., and Dennis C. Glennon, "CRA as 'market development' or 'tax': An analysis of lending decisions and economic development," Proceedings of a Conference on Bank Structure and Competition, Chicago: Federal Reserve Bank of Chicago, 1996 forthcoming, pp. 556-585.

Black, Harold, M. Cary Collins, and Ken B. Cyree, "Do black-owned banks discriminate against black borrowers?" Journal of Financial Services Research, Vol. 11, 1997 forthcoming.

Black, Harold, Robert L. Schweitzer, and Lewis Mandell, "Discrimination in mortgage lending," American Economic Review, Vol. 68, May 1978, pp. 186-191.

Board of Governors of the Federal Reserve System, "Regulation BB, community reinvestment," Federal Reserve Regulatory. Services Handbook, Vol 3, 1995, section 228.

-----, "Processing procedures for Home Mortgage Disclosure Act," technical memo, No. 96, 1995.

Boorman, J. T., and Myron L. Kwast, "The startup experience of minority-owned commercial banks: A comparative analysis," Journal of Finance, Vol. 29, 1974, pp. 1123-1142.

Bradbury, Katherine, Karl E. Case, and Constance R. Dunham, "Geographic patterns of mortgage lending in Boston, 1982-87," New England Economic Review, Federal Reserve Bank of Boston, September/October 1989, pp. 3-30.

Brimmer, A.I., "The black banks: An assessment of performance and prospects," Journal of Finance, Vol. 26, 1971, pp. 379-405.

Browne, Lynn Elaine, and Geoffrey M. B. Tootell, "Mortgage lending in Boston - A response to the critics," New England Economic Review, Federal Reserve Bank of Boston, September/October 1995, pp. 53-78.

Calomiris, Charles W., Charles M. Kahn, and Stanley D. Longhofer, "Housing-finance intervention and private incentives: Helping minorities and the poor," Journal of Money, Credit, and Banking, Vol. 26, August 1994, pp. 634-674.

Canner, Glenn B., "Redlining: Research and federal legislative response," Board of Governors of the Federal Reserve System, staff study, No. 121, 1982.

Canner, Glenn B., Stuart A. Gabriel, and J. M. Woolley, "Default risk and mortgage redlining: A study of the FHA and conventional loan markets," Southern Economic Journal, Vol. 58, July 1991, pp. 249-262.

Canner, Glenn B., and Wayne Passmore, "Residential lending to low-income and minority families: Evidence from the 1992 HMDA data," Federal Reserve Bulletin, Vol. 80, February 1994, pp. 79-108.

-----, "Home purchase lending in low-income neighborhoods and to low-income borrowers," Federal Reserve Bulletin, Vol. 81, February 1995a, pp. 71-103.

-----, "Credit risk and the provision of mortgages to lower-income and minority home buyers," Federal Reserve Bulletin, Vol. 81, November 1995b, pp. 989-1016.

-----, "The relative profitability of commercial banks active in lending in lower-income neighborhoods and to lower-income borrowers," Proceedings of a Conference on Bank Structure and Competition, Chicago: Federal Reserve Bank of Chicago, 1996 forthcoming, pp. 531-555.

Carr, James H., and Isaac F. Megbolugbe, "The Federal Reserve Bank of Boston study on mortgage lending revisited," Journal of Housing Research, Vol. 4, 1993, pp. 277-314.

Clair, Robert, "The performance of black-owned banks in their primary market areas," Economic Review, Federal Reserve Bank of Dallas, 1988, pp. 11-20.

Coplan, Stephen, "HUD offers minority lending data on Internet," American Banker, August 6, 1996, p. 11.

Dahl, Drew, "Ownership changes and lending at minority banks: A note," Journal of Banking and Finance, Vol. 20, August 1996, pp. 1288-1301.

Day, Ted, and Stanley J, Liebowitz, "Mortgages, minorities, and HMDA," University of Texas at Dallas, unpublished manuscript, March 18, 1996.

Dedman, William, "The color of money," Atlanta Constitution, four-part series, May 1-4, 1988.

Esty, Benjamin C., "South Shore Bank: Is it the model of success for community development banks?" Proceedings of a Conference on Bank Structure and Competition, Chicago: Federal Reserve Bank of Chicago, May 1995, pp. 192-217.

Evanoff, Douglas D., and Lewis M. Segal, "Strategic responses to bank regulation: Evidence from HMDA data," Journal of Financial Services Research, Vol. 11, 1997 forthcoming.

Ferguson, Michael F., and Stephen Peters, "What constitutes evidence of discrimination?" Journal of Finance, Vol. 50, June 1995, pp. 739-748.

General Accounting Office, Community Reinvestment Act: Challenges Remain to Successfully Implement CRA, Washington: Government Printing Office, 1995.

-----, Fair Lending: Federal Oversight and Enforcement Improved but Some Challenges Remain, Washington: Government Printing Office, 1996.

Garwood, Griffith L., and Dolores S. Smith, "The Community Reinvestment Act: Evolution and current issues," Federal Reserve Bulletin, Vol. 79, April 1993, pp. 251-267.

Glennon, Dennis, and Mitchell Stengel, "An evaluation of the Federal Reserve Bank of Boston's study of racial discrimination in mortgage lending," Comptroller of the Currency, working paper, No. 94-2, 1994.

Holmes, Andrew, and Paul Horvitz, "Mortgage redlining: Race, risk, and demand," Journal of Finance, Vol. 49, March 1994, pp. 81-99.

Horne, David K., "Evaluating the role of race in mortgage lending," FDIC Banking Review, Vol. 7, 1994, pp. 1-15.

-----, "Mortgage lending, race, and model specification," Journal of Financial Services Research, Vol. 11, 1997 forthcoming.

Hula, Richard C., "Neighborhood development and local credit markets," Urban Affairs Quarterly, Vol. 27, 1991, pp. 249-267.

Hunter, William C., and Mary Beth Walker, "The cultural affinity hypothesis and mortgage lending decisions," Journal of Real Estate Finance and Economics, Vol. 13, 1996, pp. 57-70.

Hutchinson, Peter M., James R. Ostas, and J. David Reed, "A survey and comparison of redlining influences in urban mortgage lending markets," AREUEA Journal, Vol. 5, 1977, pp. 463-472.

Interagency Regulatory Task Force, "Policy statement on discrimination in lending," Federal Register, April 15, 1994, pp. 18266-18274.

Kwast, Myron L., "New minority-owned commercial banks - A statistical analysis," Journal of Bank Research, Vol. 12, 1981, pp. 37-45.

Kwast, Myron L., and Harold A. Black, "An analysis of the behavior of mature black-owned commercial banks," Journal of Economics and Business, Vol. 35, 1983, pp. 41-54.

Lacker, Jeffrey M., "Neighborhoods and banking," Economic Quarterly, Federal Reserve Bank of Richmond, Spring 1995, pp. 13-37.

Lash, Nicholas, and Larry R. Mote, "Early evidence on banking in low- and moderate-income minority areas: A case study," paper presented at the Western Economics Association, Vancouver British Columbia, June 29, 1994.

Liebowitz, Stanley J., "A study that deserves no credit," Wall Street Journal, September 1, 1993.

Lindsey, Lawrence B., "Home ownership opportunities in a deregulatory environment," Marketwise, Federal Reserve Bank of Richmond, No. 1, 1996, pp. 20-27.

Macey, Jonathan R., and Geoffrey P. Miller, "The Community Reinvestment Act: Economic analysis," Virginia Law Review, Vol. 79, March 1993, pp. 291-348.

Malmquist, David, Clifford Rossi, and Fred Phillips-Patrick, "The economics of lending to low-income mortgages," Journal of Financial Services Research, Vol. 11, 1997 forthcoming.

Megbolugbe, Isaac F., and Man Cho, "An empirical analysis of metropolitan housing and mortgage markets," Journal of Housing Research, Vol. 4, 1993, pp. 191-223.

Munnell, Alicia H., Geoffrey M. B. Tootell, Lynn E. Browne, and James McEneaney, "Mortgage lending in Boston: Interpreting HMDA data," Federal Reserve Bank of Boston, working paper, No. 92-7, 1992.

-----, "Mortgage lending in Boston: Interpreting HMDA data," American Economic Review, Vol. 86, March 1996, pp. 25-54.

Perle, Eugene D., Kathryn Lynch, and Jeffrey Horner, "Model specification and local mortgage market behavior," Journal of Housing Research, Vol. 4, 1993, pp. 225-243.

Rosenblatt, Eric, "A reconsideration of discrimination in mortgage underwriting with data from a national mortgage bank," Journal of Financial Services Research, Vol. 11, 1997 forthcoming.

Ross, Stephen L., "Flaws in the use of loan defaults to test for mortgage discrimination and FHA loan performance," Cityscape, Vol. 2, February 1996, pp. 41-48.

Schill, Michael H., and Susan M. Wachter, "A tale of two cities: Racial and ethnic geographic disparities in home mortgage lending in Boston and Philadelphia," Journal of Housing Research, Vol. 4, 1993, pp. 245-275.

Seiberg, Jaret, "Efforts to halt mortgage bias seen finally bearing fruit," American Banker, July 31, 1996a, p. 1.

-----, "HMDA: Five years later," American Banker, 4-part series, September 16-20, 1996b.

Shlay, Anne B., "Not in that neighborhood: The effects of population and housing on the distribution of mortgage finance within the Chicago SMSA," Social Science Research, Vol. 17, 1988, pp. 137-163.

-----, "Financing community: Methods for assessing residential credit disparities, market barriers, and institutional reinvestment," Journal of Urban Affairs, Vol. 11, 1989, pp. 201-223.

Stengel, Mitchell, and Dennis Glennon, "Evaluating statistical models of mortgage lending discrimination: A bank-specific analysis," Comptroller of the Currency, working paper, No. 95-3, 1995.

Tootell, Geoffrey, "Defaults, denials, and discrimination in mortgage lending," New England Economic Review, Federal Reserve Bank of Boston, September/October 1993, pp. 45-51.

U.S. Congress, Senate Committee on Banking, Housing, and Urban Affairs, "Discrimination in home mortgage lending," hearing before the 101st Congress, first session, Washington: Government Printing Office, October 24, 1989, p. 1.

White, Lawrence J., "The Community Reinvestment Act: Good intentions headed in the wrong direction," Fordham Urban Law Journal, Vol. 20, November 1993, pp. 281-292.

Wilke, John R., "Mortgage lending to minorities shows a sharp 1994 increase," American Banker, February 13, 1996, p. 1.

Yezer, Anthony M. J., Fair Lending Analysis: A Compendium of Essays on the Use of Statistics, Washington: American Bankers Association, 1995.

Yezer, Anthony M.J., Robert F. Phillips, and Robert P. Trost, "Bias in estimates of discrimination and default in mortgage lending: The effects of simultaneity and self-selection," Journal of Real Estate Finance and Economics, Vol. 9, November 1994, pp. 197-215.

Zandi, Mark, "Boston Fed's bias study was deeply flawed," American Banker, August 19, 1993, p. 13.

Douglas D. Evanoff and Lewis M. Segal are economists at the Federal Reserve Bank of Chicago. They would like to thank John Bergstrom, Raphael Bostic, Paul Calem, Glenn Canner, and Lorrie Woos for helpful comments on earlier drafts, and Pat Dykes for her generous data support. They also acknowledge the technical assistance of Jonathan Siegel.