Lending discrimination in the United States has been an issue debated in the academic press and the political arena. (2) The issue of discrimination in mortgage lending has usually assumed that minority applicants were wrongfully denied credit by non-minority-owned lenders. However, a recent study by Black, Collins and Cyree (3) examined the lending decision of black applicants at black-owned banks. Prior to this study, it was assumed that minority-owned banks would have an affinity to provide credit for applicants of the same race as the bank's owners. The importance of this affinity has been discussed in previous papers, noting the role that black-owned banks play in serving low-income minority communities and aiding in the economic' development within the communities they serve. (4)
Black, Collins and Cyree and a related study by Bostic and Canner (5) consider the Metropolitan Statistical Area (MSA) as the relevant geographic market area of the bank from which to explore discrimination in mortgage lending. However, Black (6) suggests that it may be inappropriate to compare black-owned and white-owned banking institutions within a geographic area as large as a Metropolitan Statistical Area (MSA). He contends that comparing banks from such a large geographic area does not take into account that the banks could be operating in very different economic environments, which could cause differences in performance.
Black recommended using a more narrowly defined geographic area that is more representative of the bank's primary market area. He postulated that by using a more narrowly defined geographic area that takes into account the bank's primary market area, differences in bank performance may be eliminated. Based on Black's recommendations, Clair (7) and Lawrence (8) analyzed banks grouped for comparison purposes within the same ZIP code or adjacent ZIP codes. This grouping of ZIP codes was called ZIP code clusters. Clair and Lawrence find that sizable differences in performance were eliminated after the geographic area was corrected.
The use of a more narrowly defined geographic area would seem to be beneficial in the mortgage lending discrimination literature. Previous studies assume that applicants at minority-owned and non-minority-owned banks come from similar applicant populations, but these populations may, in fact, be very different and this difference may increase as the geographic area increases in size. As a result, differences in applicant populations would generate misleading results regarding the existence of lending discrimination among banks. The use of a more narrowly defined geographic area would, therefore, be advantageous in exploring the mortgage lending relationship between minority mortgage applicants and minority-owned banks.
Another problem with previous papers is that they assume that all applicants have an equal probability of encountering discrimination. However, as Calomiris, Kahn and Longhofer note, individuals with low incomes and marginal credit histories have fewer options when applying for mortgage credit and are most likely to be discriminated against by lenders. (9) Thus, we focus on low-income applicants for whom the existence of lending discrimination is less likely to be hidden by higher credit quality applicants who have very little chance of being denied credit. Moreover, Calomiris, Kahn and Longhofer contend that lenders have less difficulty evaluating the creditworthiness of borrowers who have similar backgrounds and experiences as the bank's owners, thereby leading to fewer rejections. This relationship is called the cultural affinity hypothesis.
Our paper focuses on low-income borrowers and the issues of lending discrimination and cultural affinity. It is important to note that we cannot make definitive statements regarding the existence of lending discrimination in that we utilize the HMDA data set. These data lack critical information used when making the lending decision, such as employment stability and credit history. (10) Nevertheless, Federal regulators use the HMDA data to provide an initial indication of discrimination in mortgage lending. (11) As a consequence, we can only point to indicators of discrimination rather than actual discrimination itself.
The following section is an overview of the lending discrimination and cultural affinity literature. Section 3 discuses the construction of the sample and sample characteristics. Section 4 discuses the empirical models and results. Section 5 summarizes and concludes the paper.
Previous studies have analyzed discrimination in mortgage lending by comparing the accept/reject decisions of banks based on the ownership of the lending institution and/or the race of the applicant. (12) The results associated with these studies have been mixed. A number of problems become apparent when studying the previous research on lending discrimination, notably that prior studies do not isolate those applicants who are most likely to be discriminated against because of their marginal credit history and the lack of alternative borrowing options available to them. Applicants who have traditionally been viewed as marginal and most susceptible to discrimination are low-income individuals. Previous studies assume that all applicants have an equal probability of being discriminated against, but in reality this probability increases as the individual's income level decreases. (13)
Research on the accept/reject decision of lending institutions prior to the study conducted at the Federal Reserve Bank of Boston found little evidence of discrimination in mortgage lending. (14) The 1996 Boston Fed study found statistically significant differences in rejection rates between black and white applicants. (15) However, other researchers have found data and methodological problems when trying to replicate the Boston Fed study. (16)
The Black, Collins and Cyree study changed the focus of the lending discrimination literature by addressing the question of whether black-owned banks discriminate against black borrowers. Their results show that black-owned banks are more likely to reject loan applications from black applicants than from similarly situated white applicants. They also find that black-owned banks are more likely than are white-owned banks to reject black applicants. Black, Collins and Cyree are careful to note that their results only provide an indication of discrimination given the constraints of using the HMDA data. They also point out that their results may, in fact, be a confirmation of Yezer, Phillips and Trost who argue that banks that actively lend to minorities may yield results that falsely indicate that they are discriminating. (17)
However, Black, Collins and Cyree and previous papers match banks using the Metropolitan Statistical Area (MSA) and they make no distinction between the income level of the applicants. If we follow Clair and Lawrence's work that calls for a tighter geographic definition when making comparisons between banks, it seems reasonable that a more appropriate method to determine discrimination in mortgage lending is to compare minority-owned banks with white-owned banks located in close proximity ZIP code clusters. (18) Yezer, Phillips and Trost provide statistical justification for this method in their argument regarding false indicators of discrimination. (19) Problems with false positives may be eliminated if white-owned banks are matched with minority-owned banks by ZIP code clusters.
The cultural affinity hypothesis can be divided into two testable models: lender based affinity and borrower based affinity. Lender based cultural affinity was presented by Calomiris, Kahn and …