Measuring Housing Discrimination in a National Study: Report of a Workshop (2002)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

6 Auditing Discrimination in Underserved Communities As noted earlier, underserved communities are those portions of the housing market that are not included in the HDS sampling frame used to select test sites or newspaper advertisements. For purposes of the 2000 HDS, there are two types of underserved communities: (1) neighborhoods that are underrepresented among advertisements in large metropolitan area newspapers, and (2) smaller metropolitan areas with 25,000 to 100,000 residents. In discussing the question of how discrimination in these com- munities can be audited, participants addressed issues of racial residential concentration, racial steering, racial preferences for neighborhood racial composition, and the distinction between statistical discrimination and in- dividual incidences of discrimination. Reflecting on her own research, Nancy Denton noted an apparent middle-class bias to using a sample of advertisements to construct the au- dit. This potential bias raises issues that are both statistical and substantive. Denton linked the auditors who are recruited to perform tests to actual home seekers in underserved communities. She observed that if the defini- tion of underserved includes only communities that are missed by the meth- odology used to select advertisements, researchers have failed to recognize that these areas include poor communities that are underserved in other ways as well. They are underserved by realtors, who do not want to adver- tise them, and by banks, which do not want to provide potential home buyers with mortgages. These communities are also underserved by enti- ties that are indirectly related to housing opportunities but potentially re- 38 

AUDITING DISCRIMINATION IN UNDERSERVED COMMUNITIES 39 lated to housing choices and search patterns, such as businesses and mu- nicipal services. Transportation access and education resources are also lim- ited in these communities. The varying extent to which these communities are underserved has methodological implications. In some poor communities, no housing is advertised; these communities will be missed regardless of the sampling methodology. Similarly, housing opportunities in gated communities and some working-class communities are unadvertised. Denton also noted that some housing is advertised in non-English language newspapers because the advertiser is targeting immigrants or non-American applicants, and some housing is not advertised because the landlord does not want appli- cants of a different race. As noted earlier, these housing units may not be captured by the expanded methodology proposed for Phase II of the study. Additionally, Denton suggested that researchers should consider whether the auditors could realistically assume the identities of potential home seek- ers in underserved communities, whose members may possess characteris- tics that are difficult for an auditor from a major metropolitan area to as- sume or portray. Moreover, housing search patterns may differ across income levels, and these differences can have implications for matching auditors and assigning auditor profiles. Denton commented that housing transactions for marginally qualified and overly qualified applicants are also very different, and these differences have implications for the audit results, particularly in terms of unmeasured heterogeneity. The question arises of whether auditors are paired well enough to diminish the effect of this het- erogeneity and the potential discrepancies between auditors’ actual charac- teristics and their assigned profiles. Participants recognized that auditors are extensively trained to portray various types of home seekers and that their assigned profiles may require them to depict individuals with attributes dissimilar to their own. Denton suggested, however, that it is important to consider whether auditors are trained well enough or inherently capable of assuming identities that are beyond their scope of knowledge. She noted that these issues are particularly salient when auditors visit certain kinds of communities, such as low-income or mono-ethnic communities. Denton cautioned researchers to consider potential problems with sending auditors from fair housing agencies in larger metropolitan areas to smaller underserved areas, which are unlikely to have a fair housing agency. Additionally, she posed several questions with regard to those agencies’ re- cruitment and training methods: Can a middle-income auditor with no children effectively portray a low-income single mother? Can an employed 

40 MEASURING HOUSING DISCRIMINATION IN A NATIONAL STUDY auditor who has never received public assistance effectively portray a home seeker who has just obtained his or her first job after receiving public assis- tance for many years? Can a low- or moderate-income auditor effectively audit housing units that require very high incomes? Denton expressed her belief that there are limits to the auditors’ portrayals, and that auditor train- ing does not eliminate these limitations. An additional drawback of the current methodology for auditing underserved communities relates to potential differences in housing search methods as compared with major metropolitan areas. Prior research by Denton and others has shown that residents in underserved communities are less likely to use a real estate agent during their housing search. They may also be less likely to drive around several communities to locate a desired neighborhood or housing unit. According to Denton, audit proto- cols and auditor profiles must take these potential differences into account. In her concluding remarks, Denton suggested that researchers should consider substantive issues such as those outlined above before addressing the technical aspects of auditing and measuring discrimination. She stressed that, while it is important to develop a valid, scientifically defensible esti- mate of the extent of housing discrimination in the national market, re- searchers will be unable to derive a proper estimate if they limit the scope of audit studies to the middle of the housing market. EFFECT OF NEIGHBORHOOD RACIAL PREFERENCE ON INTERPRETATION OF AUDIT RESULTS Lawrence Bobo, Department of Sociology, Harvard University, dis- cussed his research in Los Angeles on urban inequality and the work of others in this research area. Bobo’s research assessed residents’ preferences for the racial composition of their neighborhoods. Results of the study indicate that preferences for the racial composition of neighborhoods are related to race. Bobo noted that the study did not lead to recommenda- tions on how to sample housing units or assign auditors, but that its results have implications for the interpretation of results of the HDS, particularly with regard to rates of housing discrimination. Bobo’s data indicate clearly that in the general housing market, some communities are more likely to accept or reject particular racial groups. These attitudes are held by both majority and minority residents and can have implications for the way applicants of a given race are treated in a housing market transaction. Research on racial residential segregation also 

AUDITING DISCRIMINATION IN UNDERSERVED COMMUNITIES 41 confirms that many neighborhoods are racially typified (e.g., as a Hispanic neighborhood). Such typifying affects whether certain groups will pursue housing in these neighborhoods. Bobo expressed concern that the current method of sampling advertisements does not account for these kinds of community dynamics and the interaction between racial preference and housing search patterns. He suggested that it is insufficient to randomly sample advertisements in local community-based newspapers because dif- ferent racial groups may consider varying segments of the housing market. Some racial groups may exclude housing opportunities in certain neighbor- hoods from their housing search. Research on racial residential segregation might help identify neighborhoods or communities that are not included in the housing search of certain racial groups. A participant asked about the implications of the Los Angeles study for the “tipping point” of a community—the point at which people start mov- ing out because of increases in the proportion of minority residents—and the relationship to housing availability. Neighborhood racial preference could affect the vacancy rates in particular neighborhoods. Specifically, the sample of advertisements may include a higher proportion of mixed or racially transitional neighborhoods as majority households move out be- cause of increases in the proportion of minorities. The sampling method- ology may miss stable all-white or all-minority neighborhoods where there is less movement. In response to Bobo’s comments, Margery Turner of the Urban Insti- tute stated that empirical experience from the 1989 HDS suggests that minority communities, especially those in the central city, are underrepre- sented in the HDS newspaper advertisement sample. It is not known whether protected white communities in the suburbs are also underrepre- sented. It is clear, however, that middle- and high-income minority com- munities were underrepresented in the 1989 sample of advertisements. Participants discussed the need for data on the turnover rate for rental and sales housing for both racially stable and transitional communities. While researchers can obtain information on housing stock, basic turnover rate data do not exist on a national level. A participant asked whether the 1989 HDS provided some evidence that housing agents advertise units they are willing to show to anyone, regardless of race. This behavior would result in a lower incidence of racial steering. Turner responded that because minority and mixed neighbor- hoods were underrepresented in the 1989 newspaper advertisement sample, minority auditors were generally shown housing units in white neighbor- 

42 MEASURING HOUSING DISCRIMINATION IN A NATIONAL STUDY hoods. Researchers noted that in 1989 there was some steering to lower- value neighborhoods, but not much steering to neighborhoods with a higher proportion of minorities, because the latter were underrepresented in the sample. RACIAL STEERING Bobo noted that the existence of racial steering has implications for the current HDS study design and interpretation. If real estate agents have, and act upon, assumptions about housing seekers’ racial preferences in resi- dence, individual home seekers will not be shown housing units in certain neighborhoods as a result. A participant presented an alternative motivation for racial steering by housing providers, suggesting they may be motivated by profit maximiza- tion rather than racial prejudice or agents’ perceptions of client or commu- nity preferences. Housing providers may be more likely to steer white customers to white neighborhoods because they think doing so will mini- mize the amount of time it takes to fill vacant units. Housing agents oper- ating in this manner will be less worried about the preferences of minority applicants. They will, however, be concerned with the effect of renting or selling to an African American customer on their current or potential cli- ents in white neighborhoods. Evidence of this behavior may be found in steering of white households away from minority neighborhoods or steer- ing of minority households away from white neighborhoods. Bobo noted it is not clear that this form of racial steering hurts whites, unless one takes a broader view of discrimination and its general effect on society. According to Bobo, the above processes are confirmed by respondents’ views on neighborhood desirability as reported in the Los Angeles area study. Many white residents considered affluent majority African Ameri- can communities to be less desirable than lower-valued white communi- ties. Community-held views of neighborhood racial composition may therefore propagate racial residential segregation. These results suggest that the random selection of advertisements from newspapers does not account for neighborhood self-selection exhibited by actual home seekers. Bobo commented further that the steering of white households may actually hurt African American homeowners by changing the demand for their housing. Urban Institute researchers cited evidence from the 1989 HDS that a requested housing unit’s characteristics and location served as a signal to real estate agents and housing providers with regard to the type of 

AUDITING DISCRIMINATION IN UNDERSERVED COMMUNITIES 43 neighborhood to which the requester should be directed. The researchers believe it is important to consider the broad methodology used for picking an advertised unit and selecting two auditors who are requesting that unit. Requesting an advertised unit appears to influence behavior in a specific way: if an auditor asks for a unit in a certain type of neighborhood, he or she will be more likely to view additional units in that type of neighbor- hood. Real estate agents are prompted by any information they can obtain to serve their customer, but they do not apply that information in a race- neutral manner. Bobo noted that observed steering behaviors based on requests for advertised units vary by race. ISSUES SPECIFIC TO ASIAN AMERICAN AND AMERICAN INDIAN POPULATIONS As noted in Chapter 1, an important, new goal of the 2000 HDS is to develop estimates of housing discrimination for Asian Americans and American Indians. Workshop participants addressed the potential impact of testing involving these populations. Min Zhou, Office of Educational Research and Improvement, U.S. Department of Education, offered some comments about the study design and substantive issues related to auditing within the Asian American and American Indian communities. She noted that the potential bias against immigrants has implications for audits performed in metropolitan areas. Moreover, metropolitan areas with high proportions of immigrant residents often have a different housing market structure from that of other metro- politan areas. For example, minority immigrants tend to concentrate in certain neighborhoods and to have their own housing market. As a result, there may be several housing submarkets operating within a metropolitan area: (1) exclusively majority, (2) exclusively African American or His- panic, (3) mixed or “open,” and (4) exclusively Asian American. In addi- tion, Asian ethnic groups further segment the latter housing market. The dynamics of these housing markets are different from those of the general housing market. Zhou explained that real estate agencies are a very important part of the Chinese and Korean ethnic economy and that they tend to target par- ticular ethnic groups in advertising. Advertising patterns within the Asian American submarkets suggest there may also be discrimination against other ethnic and racial groups. Asian Americans locate available housing by speaking with other members of their ethnic group or reading ethnic news- 

44 MEASURING HOUSING DISCRIMINATION IN A NATIONAL STUDY papers. Since these newspapers are written in Chinese or Korean, respond- ing to their advertisements may not represent a realistic point of entry into the market for auditors from other ethnic groups. Zhou stressed that re- searchers must recognize these alternative points of entry because they are where a substantial proportion of inquiries by Asian American home seek- ers begin. The current HDS newspaper sampling methodology would miss these sources, but including them in the expanded sampling frame might not be appropriate. Thus, according to Zhou, there are portions of the housing market that are inaccessible by the audit study design. Zhou suggested expanding the concept of discrimination. The current study estimates mainly discrimination by whites against African Ameri- cans, Hispanics, Asian Americans, and American Indians. It does not ex- plore the pattern of discrimination exhibited by Asian or Hispanic housing providers. Zhou’s definition of discrimination is more varied because it includes Asians discriminating against Asians, Hispanics, and African Americans. In terms of white discrimination against Asian Americans, Zhou ob- served that Asian Americans who speak with an accent can be viewed as foreign and treated on the basis of stereotypes associated with foreigners. She used as an example the stereotype that immigrant Asians have the fi- nancial resources to purchase housing with cash or make a substantial down payment. In addition, there are negative stereotypes associated with work- ing-class Asians and perceived differences in lifestyle. Thus, the stereotypes applied may be positive or negative and may result in differential behavior by the housing agent. Zhou observed that many Asian Americans are unfamiliar with fair housing laws and are not aware of their rights under the Fair Housing Act. She noted there is substantial anecdotal evidence of racial steering for Asian American households. This steering is carried out by housing providers, as well as friends and family of the home seeker, and is perceived as being helpful. Zhou also cited the increasing tendency of real estate agencies to hire Asian or Hispanic agents. It is unclear, however, whether the objective is to systematically steer or to legitimately assist home seekers. The amount of money applicants are asked to provide for a down payment or security deposit is another example of housing market dis- crimination against Asian Americans. While overall mortgage denial rates may be lower for Asian Americans than for other minority groups, Asian Americans may pay a higher proportional down payment. This larger per- 

AUDITING DISCRIMINATION IN UNDERSERVED COMMUNITIES 45 centage will increase the chances that the mortgage will be originated, but may not be viewed as discrimination when analyzing market-level data. A participant asked whether the extent of heterogeneity among Asian Americans prevents researchers from measuring discrimination in that group. Zhou responded that the HDS does not include sufficient observa- tions to explore discrimination across ethnic groups within the Asian com- munity, and that case studies would be more effective for this purpose. She noted further that there is considerable diversity within the Hispanic popu- lation, but that Hispanics and African Americans may have common his- torical or cultural experiences that result in similarities in their discrimina- tion experiences. Thus differences in language, religion, and national origin among Asian populations present substantial difficulties for interpreting audit results. Responding to Zhou, Stephen Fienberg noted that her observations imply there are separate universes with distinct sampling frames from which measurements are made. Housing availability notices and housing market transactions may be structured differently in these markets and vary across test sites. To the extent that these structures are not known and sampling of advertisements is not viable, many underserved communities will be missed. Joseph Altonji suggested a way of addressing multiple listing sources for advertisements and differential access to advertisements in some sources: (1) combine advertisements from all newspaper sources—including non- English language papers, and (2) use audit results to assess which advertise- ment sources are open to all groups regardless of race. Assuming advertise- ments in papers of a certain language are open only to that ethnic group (e.g., Korean or Chinese), researchers could draw conclusions about equal access for underserved populations. Altonji added that the assumption might not be appropriate for Spanish-language papers that serve Hispanic communities because a larger proportion of the non-Hispanic population speaks Spanish than speaks Chinese or Korean. Conclusions about the penetration of various sources into underserved communities and the rea- sonableness of sending majority auditors to units advertised in ethnic pa- pers may require going beyond the auditing framework to assess discrimi- nation against particular ethnic groups. The discussion of auditing in Asian communities also addressed the question of whether the paradigm of paired testing makes sense in segre- gated housing markets that attempt to accept only Asian ethnic groups. Some participants suggested that the best way to understand this issue is within the framework of varying degrees of discrimination. In this 

46 MEASURING HOUSING DISCRIMINATION IN A NATIONAL STUDY conceptualization, there is some discrimination that is essentially benign and serves to meet the needs of populations underserved by the general housing market. This discrimination is viewed as beneficial because it is a parochial phenomenon, initiated by people who are trying to help indi- viduals who might otherwise be discriminated against by the majority group. On the other hand, some participants stressed that every act of discrimination, regardless of the initiating or benefiting group, is legally wrong. Research confirms that housing choice and household neighborhood preferences are issues within Asian communities. Focus group research associated with the Los Angeles study addressed issues of housing, employ- ment, and intergroup relations. Within the Korean focus groups, each individual found his or her job through one source—the Korean Daily News. This finding was not characteristic of any other group, including the Chi- nese focus group, whose members, like those of the Korean group, were 100 percent foreign born. The focus group results for the Los Angeles study also inform the design of the HDS audits. For populations with a large proportion of foreign-born members, the major metropolitan news- paper will not fully capture the housing dynamics faced by underserved individuals. There was a brief discussion of auditing in American Indian and rural communities. Participants considered the idea of addressing acknowledged difficulties in auditing in rural communities by linking auxiliary studies in Phase II with other data collected by the Urban Institute. The question of how American Indian communities are defined was raised. Researchers responded that the communities to be audited are not on tribal land, but in a fairly large metropolitan area and one small metropolitan area that ad- joins tribal lands. To achieve adequate coverage of available housing in these areas, the newspaper sampling frame combines up to ten rural and small-town newspapers. The audits will assess whether the basic measures of differential treatment apply in American Indian communities. The Urban Institute researchers commented on pilot studies in Ameri- can Indian communities. Pilot testing has revealed that for American Indi- ans, the definition of available housing stock must be expanded beyond the existing protocol to include manufactured housing. American Indian popu- lations tend to be concentrated in small, rural metropolitan areas in which this form of housing is more prevalent. Further, this housing type is not typically advertised in major metropolitan newspapers. Researchers plan to consult with the local fair housing agency to find a point of entry into these 

AUDITING DISCRIMINATION IN UNDERSERVED COMMUNITIES 47 underserved communities. Additional sources for available housing stock are community newspapers and postings in community centers or on the Internet. Results from the pilot studies will provide substantial information about how to sample beyond a single metropolitan newspaper in small communities and adjoining rural counties. Researchers also expect to learn a great deal about whether and how to recruit American Indians as audi- tors. The pilot studies will be informative as well about the feasibility of sending white auditors into adjoining counties that have a high proportion of American Indian households. Results from the smaller pilot studies will be used to determine the feasibility of replicating the study of these com- munities on a larger scale. Participants asked whether the underserved communities would be analyzed separately given the number of audits performed. The audit re- port will include simple comparisons and will address differences in pat- terns of discrimination and the existence of a racial dimension to those differences. Participants also inquired about the extent to which HUD is interested in alternative methodology that could make it possible to esti- mate discrimination in underserved communities and provide supplemen- tary information for the HDS on the relationship between race and hous- ing search patterns, as well as other housing market characteristics. In addition, participants discussed the importance of research studies address- ing the identification of an unbiased point of entry into the market that would allow for comparisons, measurement, and analysis of issues related to the sampling design and model estimates and other aspects of the study design. Housing research studies focused on these issues could lead to improvements in the HDS audit design.