Federal law prohibits housing discrimination on the basis of seven protected classes including race (see Box 1-1). Despite 30 years of legal prohibition under the Fair Housing Act, however, there is evidence of continuing discrimination in American housing, as documented by several recent reports (e.g., Massey and Lundy, 1998; Ondrich et al., 2000). In 1998, the Department of Housing and Urban Development (HUD) funded a $7.5 million independently conducted Housing Discrimination Survey (HDS) of racial and ethnic discrimination in housing rental, sales, and lending markets (Public Law 105-276). This survey is the third such effort sponsored by HUD. Its intent is to provide a detailed understanding of the patterns of discrimination in housing nationwide.
In 1999, the Committee on National Statistics (CNSTAT) of the National Research Council (NRC) was asked to review the research design and analysis plan for the 2000 HDS and to offer suggestions about appropriate sampling and analysis procedures. The review took the form of a workshop that addressed HUD's concerns about the adequacy of the sample design and analysis plan, as well as questions related to the measurement of various aspects of discrimination and issues that might bias the results obtained. The discussion also explored alternative methodologies and research needs. In addition to addressing methodological and substantive issues related specifically to the HDS, the workshop examined broader questions related to the measurement of discrimination. The workshop participants (listed in Appendix C) included representatives of
BOX 1-1 Overview of the Fair Housing Act
The 1968 Fair Housing Act prohibits discrimination in housing on the basis of five protected classes: race, color, religion, sex, and national origin. The act was amended in 1988 to expand the protected classes to include familial status and handicap. Individual jurisdictions may add to but may not subtract from the seven federally protected classes.
U.S. Code Title 42, Chapter 45, states the provisions of the Fair Housing Act. With the exception of exempted units, it is unlawful to engage in “discriminatory housing practices” in the sale or rental of a housing unit. The code outlines prohibited practices, which include refusing to sell or rent a unit and offering different terms or conditions on the basis of any protected class.
Section 3603 of the code allows for certain exemptions. One is for the lease or sale of a single-family house by the owner if that individual owns no more than three single-family houses. Additional provisions obtain, including that the house must be sold or rented without using a real estate agent or advertisement in violation of fair housing codes. Also exempted are units in living quarters occupied by the owner, provided no more than four families reside in the dwelling independently.
Section 3604 sets forth practices prohibited under the Fair Housing Act. Among these practices are the following:
One or more of the seven protected classes cannot be a factor in the outcome of the housing transaction.
HUD and the Urban Institute (which is conducting the survey), as well as experts in the salient methodological and substantive areas. Formal recommendations were not developed because the discussion of these issues constituted the review required by HUD. This report provides a summary of the workshop discussions.
In addition to informing HUD's plans for the HDS, the workshop served as preparation for an upcoming NRC project on methods for assessing discrimination. This study will address broader methodological approaches for defining and measuring discrimination, incorporating what has been learned from the housing field and audit studies. It will examine the range of methods in current use and produce recommendations regarding those which most reliably differentiate discrimination from other differences. Both projects are part of a larger body of DBASSE reports on subjects related to the achievement of equal opportunity. 1
THE HOUSING DISCRIMINATION SURVEY
The HDS is being conducted in three phases. The first phase, which began in 2000, involved an initial set of 20 metropolitan areas and as many as 6,000 matched-pair audits. The initial 20 sites allowed for over-time comparisons with the results of the earlier two surveys. An important aspect of the design is the determination of an “optimum” selection of an additional 40 sites to generate a valid national estimate of discrimination against specific minority groups (African Americans, Hispanics, Asians, and American Indians). As noted, the survey is being conducted by the Urban Institute, which also carried out the two previous discrimination surveys. A detailed description of the three phases of the HDS is provided in Chapter 2.
The HDS is using matched-pair audits to investigate housing discrimination, building on HUD's 20-year reliance on this methodology. In these
1A Common Destiny: Blacks and American Society (1989); Measuring Poverty: A New Approach (1995); Title I Testing and Assessment: Challenging Standards for Disadvantaged Children (1996); Effects of Welfare on the Family and Reproductive Behavior (1998); Racial and Ethnic Differences in the Health of Older Americans (1997); Demographic and Economic Impacts of Immigration (1997); Health and Adjustment of Immigrant Children and Families (1998); Improving Schooling for Language-Minority Children (1998); Improving the Future of U.S. Cities Through Improved Metropolitan Area Governance (1999); and America Becoming: Racial Trends and Their Consequences (2001).
audits, pairs of individuals are matched for “all” relevant characteristics other than those that are expected to lead to discrimination. If there are significant differences in the way members of the pairs are treated in the test setting, the matched-pair audit methodology attributes these differences to discrimination. The strengths and limitations of this approach are discussed in detail in Chapter 2.
HUD's goal is to develop statistically valid measures of the extent of racial and ethnic discrimination throughout a sample of American housing markets. The agency plans to use the results of the study to target future enforcement efforts more effectively, to direct legislative action needed to reduce discrimination, and to create “report cards” for the nation and for the communities studied that can be used to measure progress toward the goal of greater social integration.
KEY ISSUES DISCUSSED AT THE WORKSHOP
The major objective of the HDS is to measure the incidence of racial discrimination in the national housing market. The audit methodology and sampling frame applied to study discrimination against minorities in urban housing markets raise several key issues. The workshop discussions addressed these issues from both a methodological and substantive perspective.
Prior housing discrimination studies have focused on discrimination against African American and Hispanic households, whereas HDS 2000 extends the minority groups studied to include Asian Americans and American Indians. Consequently, an additional goal of the workshop was to address conceptual, methodological, and sociological issues related to measuring housing discrimination against these minority groups. Participants were also asked to address the application of the current audit methodology and sampling frame to studying subpopulations within the Asian community, as well as concentrations of American Indian groups in rural communities. Including these “underserved” populations raises additional measurement issues with regard to the study design, and the workshop discussions addressed these issues as well. 2
Differential treatment of minority and white home seekers in a hous
2Underserved populations or communities are those that have not previously been included in the HDS sampling frame, because of either their racial composition or their size.
ing market transaction does not necessarily mean that racial discrimination has occurred. The purpose of paired testing is to provide an objective means of determining whether differential treatment during a housing transaction is due to the race of the applicant. Since “all” household and housing unit factors other than race are controlled, racial discrimination is assumed to be the reason if differential treatment occurs. 3 This conclusion is based on the definition of racial discrimination as “the unequal treatment of equals on the basis of race” (Fix et al., 1993). The audit test also seeks to establish a realistic point of entry into the housing market and to control for various observable factors. Of course, the methodology does not control for all factors and is based on some untested assumptions. Housing market transactions, for example, involve both random and systematic factors. Characteristics or behaviors of the auditors or housing agents, among other factors, may affect the audit results. These random factors are not controlled for by the methodology and thus may not be observable by the audit researchers.
Past research addressing both systematic and random factors involved in housing transactions has led to the development of four measures of unequal or disparate treatment: (1) discriminatory inclination, (2) gross unfavorable treatment, (3) systematic unfavorable treatment, and (4) net market effects (Fix et al., 1993). The discussions at the workshop focused on two of these measur—gross unfavorable treatment and net market effect. Since these two measures mean different things depending on how the population of interest is defined, and since all estimates of housing discrimination are based on numerous audits, two other key issues were discussed: the appropriateness of weighting the audit results and the need for a clear definition of the population of interest.
HUD officials and researchers are currently weighting the results of audits performed on advertisements sampled from major metropolitan newspapers to obtain a more accurate estimate of discrimination in the housing market. Yet the appropriateness of this weighting scheme is contingent on the definition of the population of interest. The sampling frame may not reflect the entire housing market, but rather those housing units that are advertised in major metropolitan newspapers. This issue is most
3The characteristics controlled for include all legitimate reasons a minority applicant might receive treatment different from that of a majority or white applicant, as well as other illegitimate or illegal reasons for denial (such as familial status).
salient when one considers that majority and minority home seekers may face a dual housing market; the sources used for seeking a housing unit may also differ by race. The design of the sampling frame should ideally reflect these differences.
The workshop was organized into three sessions addressing the following topics: (1) the purpose of and key policy and methodological issues related to the HDS, (2) preparations for Phase II of the HDS audit— auditing of discrimination in underserved urban communities and implications of the preceding methodological discussion for the Phase II design plan, and (3) the HUD and other methodologies for measuring discrimination. This report is structured to reflect the key issues raised on each of these topics and their relevance to the objectives of the workshop.
Chapter 2 presents an overview of the 2000 HDS, including its objectives and design. This chapter also reviews the testing methodology, with emphasis on its application to the detection of housing discrimination. Included is a summary of the workshop discussion on the advantages and disadvantages of paired testing, as well as methodological concerns regarding its use to identify discrimination in housing markets. Chapter 3 summarizes the workshop discussion on clearly defining or identifying the population of interest, in particular on whether the current HDS sampling methodology and study design do, in fact, capture the population about which inferences are drawn by researchers. Chapter 4 presents highlights of the discussion on defining housing discrimination and on the important distinction between disparate impact discrimination and disparate treatment discrimination. Chapter 5 summarizes the discussion of how to model and define housing discrimination. Finally, Chapter 6 addresses special concerns related to applying the audit study design to underserved populations, particularly Asian Americans and American Indians and those living in small metropolitan areas. In addition, three appendices are provided: Appendix A is a paper prepared for the workshop that gives a detailed description of the HDS; Appendix B is a second paper addressing some methodological issues associated with the HDS audit in a framework that is in some ways richer than that which has spawned the paired-testing methodology; and Appendix C contains the workshop agenda and a list of the workshop participants.