Abstract
In a departure from most studies of the causes of racial residential segregation that focus on the three main factors of economics, preferences, and discrimination, this paper examines one of the mechanisms through which segregation may be perpetuated: the housing search process itself. Data come from a 2004 face-to-face survey of an area probability sample of African American and white householders living in the three counties of the Detroit metropolitan area (n=734). These data are used to address three research questions: (1) What are the strategies people use to find housing, and are there racial differences in those strategies? (2) Do whites and African Americans report similar or different experiences in the search for housing? (3) Do the locations in which people search for housing vary by race? Results show that once controlling for the type of search and background characteristics, the search strategies are generally similar for whites and blacks, though more so for buyers than renters: for example, black renters use more informal strategies and networks than do white renters. Analyses that look at the features of these strategies, however, reveal some significant racial differences. Search experiences are similar in terms of length and number of homes inspected, but other objective and subjective questions about the search show blacks at a disadvantage compared to whites: African Americans submit more offers/applications for homes, report more difficulties, and are much more likely to feel they were taken advantage of during the search. The racial characteristics of the communities in which blacks and whites search are quite different: whites mainly search in white communities, while African Americans search in communities with a variety of racial compositions. The paper concludes with a call for further research on housing search strategies, with particular attention to the role of social networks.
INTRODUCTION
Residential segregation by race persists in the United States—especially in larger and older cities in the East and Midwest such as Detroit, Chicago, Milwaukee, New York, and Newark, where levels have declined only slightly in recent decades (Logan et al. 2002). Detroit, the most segregated metropolitan area in the nation, has shown barely perceptible drops in segregation: between 1980 and 2000, the black-white dissimilarity index decreased just three points to 85 (Lewis Mumford Center 2001).
The causes of persistent patterns of racial residential segregation have been hotly debated for decades, with the focus traditionally on three factors: economics, preferences, and discrimination. The first generally utilizes census data to describe who lives where and with what income levels (e.g., Massey and Fischer 1999; Darden and Kamel 2000; Alba et al. 2000; St. John and Clymer 1999); the second typically draws on individual-level survey data about hypothetical neighborhood preferences (e.g., Farley et al. 1976; Farley et al. 1994; Krysan and Farley 2002; Charles 2006); and the third conducts audit studies to test whether white and African American experimental subjects are treated the same or differently by real estate agents and landlords (e.g., Turner et al. 2002; Massey and Fischer 2004; Massey and Lundy 2001).
Most recently, a new kind of study on the causes of segregation has emerged: mobility studies that use individual level panel survey data. These have enabled tests of hypotheses about what influences who moves and where they move to, which are derived in part from theories about the causes of segregation. As a result, rather than making inferences about the causes of segregation by examining changes in neighborhood racial composition, these studies explicitly test individual-level predictors of behaviors, and thus clarify the impact of a neighborhoods’ racial composition—above and beyond other characteristics—on moving out and moving in behavior (e.g., South and Deane 1993; South and Crowder 1998; Alba and Logan 1993).
These kinds of studies provide methodologically sophisticated analyses of the migration patterns that contribute to segregation. For example, Crowder (2000) used this approach to directly test the “white flight” hypothesis. After controlling for metropolitan, neighborhood, and individual level mobility predictors, he showed that whites were more likely to leave their neighborhood as the size of the minority population increased. Quillian (2002) used transition matrices to analyze migration data that tested the three basic theories of the causes of segregation. He identified two key patterns that helped sustain segregation: white avoidance of neighborhoods with more than a few blacks (both by exiting from, and not moving into), and a lower likelihood of blacks moving into white neighborhoods.1 Quillian (2002) points out that the latter could be a function of at least two things:
The fact that African Americans move into white neighborhoods much less often than whites is consistent with the theory that there are barriers to the mobility of African Americans into white neighborhoods, although this result could also be consistent with the possibility that blacks do not want to be racially isolated in all-White neighborhoods (p. 220).
This new generation of mobility behavior studies means we now know in more detail where individuals end up moving—both to and from—as well as the impact (or lack thereof) of a variety of individual, neighborhood, and metropolitan characteristics on such moves. But as Quillian (2002) notes, we know little about how respondents end up where they do. For example, with respect to African Americans, do they consider whiter neighborhoods but meet with barriers when actually trying to move to them? Or are such neighborhoods completely off the “radar screen” of African Americans who are looking to move? Although mobility studies like these draw our attention to individual level behaviors and provide insight into the last step of the housing process—the final destination—we still know very little about the process that leads up to that point. An exception is Crowder’s (2001) recent analysis of racial disparities in the ability to turn an intention to move into an actual move. In this analysis, Crowder (2001) demonstrates the value of moving our research further back in the mobility process to better understand how particular moves occur.
Extending research on the causes of segregation to include not simply the end result of the housing search, but also the features of the search process itself, is in the spirit of Reskin’s (2003) recent call for more scrutiny of how rather than only why racial inequality happens. In short, Reskin (2003) argues for more attention to the mechanisms by which inequality is created and perpetuated. Applied to housing, Reskin’s general principle suggests that rather than asking only why residential segregation happens (e.g., economics, preferences, and/or discrimination), we would do well to also look at how people search for housing since here the mechanisms of segregation play themselves out through the strategies, experiences, and decisions people make.
This emphasis on housing searches is not meant as yet another explanation for segregation, on the order of economics, preferences, and discrimination. Rather, the goal is to draw attention to the mechanisms through which these larger forces operate. For example, one argument is that racial residential segregation is the outcome of racial differences in economic status. While studies of 1960–1980 census data refuted that hypothesis (Denton & Massey 1988; Farley 1977), more recent analyses suggest conflicting answers to this question (Massey & Fisher 1999; Darden & Kamel 2000; Alba et al. 2000; St. John & Clymer 1999; Krivo & Kaufman 1999). In her recent extensive analysis on the topic, Charles (2006) concludes that economics, though not the most important factor, are part of the answer.
How might economics operate through housing searches to explain segregation? Social class shapes the resources at one’s disposal. For example, low income individuals have less access to one source of housing information: the internet. And high income individuals may have less access to assistance from community organizations, whose services are frequently targeted to low income populations. If the internet and community organizations are more likely to market in particular kinds of areas (racially segregated ones, for example), then this may be one mechanism whereby economics influences search strategies, which in turn perpetuates segregation. People with different levels of income and education may also have more difficulty searching—they may have to submit applications to more places or be more likely to feel taken advantage of—simply because of landlords’ financial requirements for their tenants or because they are more likely to take advantage of individuals who may be less aware of their rights. Finally, social class may shape the neighborhoods one can afford to live in; so search locations may be constrained simply based on affordability. The pertinent question is whether social class can explain away observed racial differences in these dimensions of the housing search or whether race continues to be an important factor. Once blacks and whites of similar economic standing are compared, are they equally likely to use the internet? Experience hassles and difficulties in the search? Search in the same kinds of neighborhoods?
A second explanation is that segregation is the outcome of discriminatory practices in the housing market (e.g., Munnell et al. 1996; Ross & Yinger 2002; Turner et al. 2002) that keep minorities out of certain neighborhoods. Again, we can consider how discrimination may operate through dimensions of the housing search to perpetuate segregation. Discriminatory behavior by landlords and realtors may translate into lengthier searches that are filled with more hassles and require more resources in order to be successful. Or certain neighborhoods may be perceived as hostile to African Americans—ones in which they are likely to experience discrimination—and so search locations may exclude such communities. African Americans concerned about steering or other negative treatment may avoid working with real estate agents who are white or avoid working with real estate agents altogether. And African Americans may draw more heavily on information from friends and relatives in their search for housing, as a way to gather information so as to avoid communities known to be hostile to African Americans, or to avoid encountering discriminatory landlords. To the degree that these decisions decrease the variety of neighborhoods considered, the differential use of resources may contribute to housing decisions that in turn perpetuate segregation.
A third key explanation for segregation emphasizes preferences: in essence, people live in segregated areas because they choose to do so. A community’s racial (racial composition, racial climate, etc.) and non-racial (location, property values, amenities, etc.) features can contribute to people’s preferences. This article focuses in particular on the former. Racial residential preferences themselves may be shaped by experiences with discrimination: African Americans may avoid all-white communities out of concerns for potential discrimination in such communities. In addition, preferences about racial composition can influence the variety of communities a person considers. What we know about racial residential preferences suggests that whites will search largely in communities where they are in the strong majority; and African Americans will search largely in communities where they are in the slight numerical majority, or in 50–50 neighborhoods. Finally, racial residential preferences may shape where people get their information during their housing search. Newburger (1995), for example, shows that open houses are less likely to be held in racially mixed and black communities; thus, African Americans may use this resource less often, if they have restricted their searches to these kinds of communities.
The analysis reported in this paper represents a first step toward addressing these larger questions of how individual level housing search strategies, experiences, and decisions might, in the aggregate, serve to perpetuate or attenuate patterns of racial inequality in housing. Of the three main causes of segregation, this study can provide some tests of the first—that is, an examination of the effect of an individual’s race versus social class characteristics on search strategies and experiences. However, data constraints prohibit explicit tests of the kinds of broad relationships just outlined with respect to preferences and discrimination.2 Moreover, it is beyond the scope of this paper to test a fuller model to evaluate the exact impact of observed racial differences in housing search experiences on racial residential segregation itself. Rather, this paper provides answers to three initial and critical questions: (1) What are the strategies people use to find housing, and are there racial differences in those strategies? (2) Do whites and African Americans report similar or different experiences in the search for housing? and (3) Do the racial characteristics of the locations in which people search for housing vary by race?
BACKGROUND
The search for housing is a complex and fluid process, involving a number of features, people, and institutions. Clark and Flowerdew (1982) identified five housing search dimensions: (1) duration of the search; (2) number and type of information sources used; (3) number of neighborhoods searched; (4) number of houses examined; and (5) radius of the area searched. These dimensions form the foundation of our description of housing searches. Next is a review of the existing research related either explicitly or implicitly to the three research questions.
Research Question #1: What strategies do people use to search for housing, and are there racial differences in those strategies?
There are just a few studies that have analyzed racial differences in the resources and strategies people use when trying to find a home. For example, Newburger (1995) studied recent homebuyers the Boston area and found no difference between African Americans and whites in their use of real estate agents or looking for “for sale” signs. However, blacks were more likely to consult newspaper ads and were much less likely to attend open houses than whites. In a parallel study of how homes are marketed, Newburger (1995) found that homes for sale in racially mixed or predominantly black neighborhoods were substantially less likely to hold open houses. She attributes the lower use of open houses among African Americans to the scarcity of such options in neighborhoods where African Americans are more likely to search.
Farley (1996), in an analysis of Detroiters, reports significant racial differences in the use of real estate brokers/agents, newspaper ads, for sale/rent signs, talking with friends/relatives, and using community organizations. In each case except for real estate agents, African Americans were more likely than whites to use the strategy. However, after controlling for whether the searcher was looking for a place to rent or to buy, the racial differences fell out of statistical significance with one exception: black renters continued to be significantly more likely than white renters to talk to friends/relatives in order to find an apartment (Farley 1996).
The existing studies of racial differences in housing search strategies are therefore somewhat inconsistent on the question of racial differences. One the one hand, Farley (1996) finds that racial differences largely disappear once type of search (buyer versus renter) is controlled—thus highlighting the role of social class (if we treat buyer/renter status as a partial proxy for social class) in minimizing racial differences in search strategies. The strength of this study was its inclusion of both renters and buyers, and that its sample was representative of the Detroit metropolitan area. But the analysis was limited (due to data constraints) because it could not move beyond the general type of information source/strategy (newspapers, real estate agents, etc.). Unanswered are questions about whether there are racial differences in, for example, which newspapers were used, which real estate agents were hired, or how people found these resources. Newburger’s (1995) findings suggest the potential importance of these aspects of search strategies. She finds, for example, that while blacks and whites both used newspapers in their search, they used different newspapers (Boston Globe versus Boston Herald). And, in addition to higher rates of real estate agent use among whites, there were also racial differences in how they found the agents they used.
For its part, the Newburger (1995) study is limited because it included only homebuyers, and because it studied only those moving into a single section of Boston and an adjacent census tract. While there were good reasons for this research design, its ability to generalize even to the greater Boston metropolitan area is limited. In short, the dearth of studies on the topic means that the question of racial differences in home search strategies remains a largely unanswered one.
Research Question #2: Do whites and African Americans report different experiences when they search for housing?
For the most part, the evidence about the different experiences that whites and blacks have when conducting a housing search comes from audit studies in which paired testers inquire about a particular housing unit using a proscribed strategy (phoning a newspaper listing or making inquiring at a real estate office about an advertised unit). These studies find that blacks compared to whites are shown fewer units, told about fewer available houses, less likely to have phone calls returned, and generally face poorer treatment by the real estate professionals or landlords with whom they come into contact (Turner et al. 2002; Yinger 1995; Massey and Fischer 2004; Massey and Lundy 2001). These audit studies, with their tight experimental controls, are a powerful tool for documenting if and how the experience of searching for housing can differ for blacks and whites, net of economic background.
However, audit studies do not afford information about how representative samples of whites and blacks experience their searches—searches that may or may not involve the strategy tested by housing audits. Audit studies are criticized because they test only homes that are advertised in newspapers (Galster 1992) and searchers who make “cold calls” on agents or landlords. While these studies show persistent housing discrimination, they leave unanswered questions about how searches are actually conducted and what searches are like when considered as a whole, rather than as a single encounter about a single housing unit. For example, do blacks and whites search different numbers of units over the course of their search? Do they report different levels of difficulties, as might be expected if one group experiences discrimination and the other does not? And, to what extent do whites and blacks search in the manner tested through housing audits? This survey data can complement audit studies by providing data about how actual searches play out—and how they are experienced by the individuals doing the searching.
A few studies provide hints about racial differences in housing search experiences. In a study of Pittsburgh low income residents, Cronin (1982) found that minorities searched in fewer neighborhoods, saw fewer units, phoned to inquire about fewer units, searched in a more restricted geographical area, and took longer in their search than whites. In addition, Newburger (1995) found that black homebuyers inspected about 9 fewer homes than white homebuyers. She offered three possible explanations: (1) agents discriminate against blacks and show them fewer homes; (2) the lower likelihood of open houses in racially mixed or black neighborhoods restricts the number of homes blacks see; and (3) African Americans search in a more geographically restricted area, thus limiting their options. Newburger’s data do not permit a test of these speculations, but they do point to our third research question.
Research Question #3: Do the locations in which people search for housing vary by race?
Our interests here are specifically in the racial characteristics of the communities in which people actually search for housing and whether they differ by an individual’s racial background. Recent studies have provided detailed information about people’s final choices when making a move, insofar as they are related to its racial composition (e.g., Quillian 2002; South and Crowder 1998). Similarly, research provides considerable insight about what people say they are looking for in a neighborhood in terms of its racial composition—at least hypothetically (e.g., Farley et al. 1994; Krysan and Farley 2002; Charles 2006). But existing studies have at least two limitations. First, with respect to studies that document the impact of racial composition on the locations people move into, we know little about the process that resulted in that particular move. As Quillian (2002) notes, this leaves uncertainty about the process that leads to African Americans’ low levels of moving into whiter neighborhoods: Do African Americans in large numbers search in whiter neighborhoods (but meet with barriers that reduce actual moves into such places) or are their searches limited to areas where blacks predominate?
Second, we know little about whether the racial residential preferences revealed by survey data are consistent with what people actually do when they try to find housing. That is, hypothetical neighborhood studies show white preferences for heavily white neighborhoods but with some acceptance of a handful of African American neighbors—a level of tolerance that has increased somewhat over time (Farley et al. 1994). African American racial residential preferences are frequently summarized as a desire for “50–50” neighborhoods—or neighborhoods that include slightly more black residents than white residents (Charles 2000a, b). As Krysan and Farley (2002) and Charles (2000b) argue, however, the preferences may not be as inflexible as this. While blacks find evenly mixed neighborhoods the “most attractive,” a closer look at the data also reveals openness to a wide range of neighborhoods (Krysan 2002b; Krysan and Farley 2002; Charles 2000a, b). For example, Krysan (2002b) found in a study of the reported desirability of specific communities in Atlanta, Boston, Detroit and Los Angeles, that African Americans considered even some communities with small percentages of African Americans to be very desirable places to live. Unfortunately, these data did not provide evidence about actual search behaviors, so we are left uncertain as to whether African Americans would actually search in such places or simply view these communities as hypothetically desirable places to live.
Finally, given the possibility of social desirability biases in explicit survey questions on racial issues (Krysan 1998; Krysan and Couper 2003), there is reason to question whether we can take these survey responses at face value, especially for whites, who may be particularly susceptible to social desirability pressures (Krysan 1998). It is of great value to gauge what African Americans and whites actually do when they search for housing—and to understand the racial characteristics of the places they consider—as a complement to the hypothetical neighborhood preferences studies.
Although there is a substantial body of research using audit studies and hypothetical neighborhood preferences, which provide some indirect evidence of what kind of housing searches people undergo and experience, we actually have very little direct data on the question of whether blacks and white conduct searches in the same or different ways. There are few studies that have asked blacks and whites parallel questions about all five dimensions of the housing search (Clark and Flowerdew 1982) with the expressed purpose of understanding if there are racial differences. The analysis reported in this paper begins to fill in that gap. An analysis of the consequences of racial differences in search strategies is beyond the scope of this paper, but the hope is that a detailed look at racial similarities and differences in housing searches will point to new research areas that can begin to answer the question of “how” segregation is created and perpetuated.
DATA AND METHODS
This analysis is based on data from the 2004 Detroit Area Study (Farley, Krysan and Couper 2004). The Detroit Area Study (DAS) is a multi-stage area probability sample of adults 21 years and older living in households in the three counties of Wayne, Oakland, and Macomb. Interviews were conducted from April through October 2004 and the survey achieved a 56 percent overall response rate.3 African Americans were over-sampled, as were residents of racially mixed neighborhoods. The overall sample size was 734 respondents. This study focuses on the 63 percent of black and white respondents who had searched for housing in the past 10 years and were involved at least minimally in that search.4 The survey was conducted primarily as a computer assisted personal interview, although the questions on search locations (as further described below) were asked using a paper and pencil exercise.
Measures for Research Question #1: Strategies used in the housing search
To measure the strategies Detroiters used to conduct their housing search, respondents were presented with a list of 12 different resources (talking to friends, talking to relatives, newspaper, internet, signs in the yard, open houses, real estate agents, apartment locator services, rental/property management firms, community or church organizations, attending a homebuyer’s seminar, and receiving information from an employer) and asked to identify all of the strategies they had used in their most recent search. Respondents were then asked to select the one they used “the most.” They were also asked (if relevant) how they found their real estate agent and mortgage broker,5 the race and gender of their real estate agent and, if they used the internet, how they used it.
Measures for Research Question #2: Experiences in the housing search
Respondents were asked both subjective and objective questions about their most recent housing search, including how long the search took, how many homes they inspected during the search, and how many offers they tendered (for buyers) or applications they submitted (for renters). Three subjective measures were included: (a) a rating scale of how easy or difficult it was to get a mortgage (1=very easy; 5=very difficult); (b) a rating scale of how difficult, overall, the housing search was (1=extremely difficult; 5=not at all difficult); and (c) a yes/no question about whether they felt they had been taken advantage of at any time during the search.
Measures for Research Question #3: Search Locations
Our study included an innovative measure of actual housing search locations. Respondents were given an 11″×17″ spiral bound booklet of color maps that identified thirty-three communities. One of the maps asked respondents to mark the communities where they had looked for a house or apartment in the past 10 years. Respondents were able to efficiently and effectively work with the maps—just 9 percent of the respondents needed interviewer assistance with this task.
The 33 communities identified on the map were a subset of all communities in the Detroit metropolitan area. They were selected to be recognizable within the Tri-County area and to have diverse socio-economic and racial compositions, histories of racial animosities, and geographical proximity to the city of Detroit. Five of the areas identified are sections within the boundaries of the city of Detroit (Northwest Detroit, East Side Detroit, Southwest Detroit, Midtown Detroit and Downtown Detroit). The remaining 28 are suburbs in Wayne, Oakland and Macomb counties. Because the map identified only a subset of all possible places, it is useful to characterize the kinds of communities included on the map. Treating the city of Detroit as a single Census Designated Place (CDP), as the Census Bureau does, our map identifies 29 distinct places. Of these 29 places, the vast majority—76 percent—were predominately white (80 percent or more), 3 percent were predominately black (80 percent or more), and 21 percent were racially mixed (20 percent or more of at least two racial groups).6 Using these definitions, our 29 places “over-represent” racially mixed places. Considering all CDPs in the three counties, about 88 percent are predominately white, 2 percent are predominately black, and just 9 percent are racially mixed; another 2 percent are some “other” racial mix.
It is also useful to know what kinds of communities were not included on our map since a substantial percentage of respondents—especially among whites—report that while they searched for housing in the past 10 years they had not searched in any of the 33 areas identified on the map. Ninety-one percent of the places not labeled on the map (but in the Tri-County area) are predominately white; by contrast, just 5 percent are racially mixed; 1 percent are predominately African American; and 2 percent are some “other” mixture. Thus, if a respondent reports that they have searched for housing, but not in any of the labeled locations, chances are high that they searched in a predominately white community.
Measures of control variables for multivariate analyses across all three research questions
Race of the respondent was self-reported, and individuals were allowed to select more than one racial category, though very few did (n=29). Respondents were also asked to self-report Hispanic origin. Only respondents who said they were African American and did not self-report Hispanic ethnicity were classified as non-Hispanic black. Respondents were regarded as white only if they did not select any other race/ethnicity and did not report that they were of Arab, Chaldean or Persian descent. In addition, where respondents reported an “other” racial category, and also provided a description that had recognizable racial signifiers, appropriate codes were assigned (e.g., Italian or Irish were coded as white). The sample is restricted to non-Hispanic blacks and non-Hispanic whites because there were not enough cases of other groups to provide reliable statistical tests.
Education was measured using four dummy variables, with having a college degree as the reference category, and having some college, a high school diploma, and less than a high school diploma as the remaining three categories. Type of search was measured with a dummy variable, where a “1” represented those who were searching for a place to purchase or the few who said they were searching either to rent or to buy. Life course and demographic controls include respondent’s age, gender, and the presence of children less than 18 years of age in the home.7 The models also include a measure of the number of years the respondent has lived in the Detroit metro area. Income was measured by four dummy variables (with a reference category of $80,000 or more annual family income). Stata’s IMPUTE procedure was used to impute missing data on income since it was the only variable with much missing data (10%).8
The general analytic strategy consists of bivariate analyses to assess overall effect of respondent race, followed by linear, logistic, or multinomial logistic regression models, depending on the structure of the dependent variable. Where models are presented, the first generally includes respondent race as the sole independent variable. Subsequent models include social, economic, and demographic variables that may influence housing searches; these are largely used as control variables to assess whether such differences will explain (and perhaps eliminate) observed racial differences.
All analyses use the SVY command in Stata to adjust the standard errors in light of the sample design, which was both clustered and stratified. In addition, data were weighted to adjust for differential probability of selection across strata as well as within-segment nonresponse. The analysis includes individuals who had searched for housing in the last 10 years, regardless of whether that search resulted in a move to their current home. That is, eighteen percent of the respondents had not moved in the past 10 years, but had nevertheless undergone a housing search at some point during the preceding 10 years. Regardless of whether they moved or not, all searchers are included in the analyses reported in the tables in this article. However, separate analyses were conducted that limited the sample to only those who had both searched and moved in the past 10 years. Where conclusions differ for this subset of “successful searchers,” they are reported in the text or footnotes. 9
RESULTS
Research Question #1: What strategies do people use to search for housing, and are there racial differences in these strategies?
Overall Patterns in Search Strategies
The first column in Table 1 shows that, on average, respondents used just about three different strategies in their housing search, with real estate agents used by the highest percentage of searchers, followed by looking for signs in yards, reading newspaper ads, and talking to friends. Very few respondents used apartment locator services, property management firm/rental agents, or community/church organizations. Just 1 percent of searchers attended homebuyer’s seminars or had employers who provided information to them and because of this small number these two strategies are dropped in the remaining analyses.
Table 1.
Search strategies used in most recent search, overall and by renter/buyer status, 2004 Detroit Area Study.
| Total | Renters | Buyers | p-value* | |
|---|---|---|---|---|
| Strategy | ||||
| Mean number of strategies used | 2.8 | 2.6 | 2.8 | n.s. |
| Networks | ||||
| Talk to friends | 35% | 36% | 34% | n.s. |
| Talk to relatives | 28% | 29% | 27% | n.s. |
| Own Research | ||||
| Newspaper | 36% | 43% | 34% | n.s. |
| Internet | 28% | 37% | 26% | n.s. |
| Signs in Yard | 44% | 27% | 49% | p<.001 |
| Open Houses | 30% | 6% | 37% | p<.001 |
| Professionals | ||||
| Real Estate Agent | 50% | 15% | 60% | p<.001 |
| Property | 8% | 19% | 5% | p<.001 |
| Management | ||||
| Firm/Rental Agent | ||||
| Community or Church | 3% | 7% | 2% | n.s. |
| Organization | ||||
| Apartment Locator | 7% | 25% | 2% | p<.001 |
| Service | ||||
| Other | ||||
| Homebuyer’s | 1% | 0% | 1% | n.s. |
| Seminar | ||||
| Employer | 1% | 2% | 1% | n.s. |
| Strategy Used Most | ||||
| Networks | 15% | 21% | 13% | |
| Own research | 46% | 50% | 45% | |
| Professionals | 36% | 20% | 40% | |
| Other | 4% | 9% | 2% | |
| Total | 100% | 100% | 100% | p<.05 |
| How Used Internet (among those using internet) | ||||
| Listings | 84% | 86% | 83% | n.s. |
| Mortgages | 36% | n/a | n/a | |
| Info. On Communities | 50% | 41% | 54% | n.s. |
Column 3 reports p-values for test of difference between buyers and renters taking into account sample design. Sample size is 415 except for the internet questions (n=93); and mortgage and internet questions (n=60).
The data bear out the assumption that search strategies will differ for buyers and renters (columns 2 and 3 in Table 1). While drawing on networks (friends and relatives) and searching the newspaper ads and the internet are used at the same rate for renters and buyers, all of the other strategies are used at significantly different levels depending on search type. For example, a majority of buyers draw on professional assistance, but many also do their own “legwork” by looking for signs in yards and attending open houses. Search strategies among renters are more diffuse, and rely far less on professional services. Indeed, no single search strategy is used by the majority of rental searchers, with newspapers, the internet, and talking to friends the most frequently used.
After respondents identified all of the strategies they used, they were asked to select the one they used “the most.” Collapsing the strategies into four general categories (networks; own research; professional assistance; and other) reveals that about one-half of both renters and buyers used their “own research” strategy the most. Beyond this, renters relied about equally on personal networks and the help of professionals, while buyers were much more likely to rely on professionals and less likely to draw on their networks “the most.”
Racial Differences in Search Strategies
Turning to the question of whether search strategies differ for African Americans and whites, as Table 2 shows, blacks and whites overall use the same number of housing search strategies. Additionally, they are equally likely to talk to friends and relatives, to use newspapers and apartment locator services, and to look for signs in yards. But whites are statistically significantly more likely to use the internet, attend open houses, and consult with real estate agents, while African Americans are more likely to use property management/rental agents and community organizations.
Table 2.
Search strategies used in most recent search by respondent race, 2004 Detroit Area Study.
| Whites | African Americans | p-value* | |
|---|---|---|---|
| Strategy | |||
| Mean number of strategies used | 2.8 | 2.7 | n.s. |
| Networks | |||
| Talk to friends | 33% | 38% | n.s. |
| Talk to relatives | 27% | 29% | n.s. |
| Own Research | |||
| Newspaper | 35% | 36% | n.s. |
| Internet | 31% | 16% | p<.01 |
| Signs in Yard | 45% | 40% | n.s. |
| Open Houses | 34% | 18% | p<.01 |
| Professionals | |||
| Real Estate Agent | 52% | 40% | p<.10 |
| Property Management | 5% | 19% | p<.01 |
| Firm/Rental Agent | |||
| Community or Church Organization | 2% | 9% | p<.05 |
| Apartment Locator Service | 6% | 9% | n.s. |
| Strategy Used Most | |||
| Networks | 13% | 22% | |
| Own research | 47 | 40 | |
| Professionals | 36 | 33 | |
| Other | 5 | 5 | |
| Total | 100% | 100% | n.s. |
| How Found Professionals | |||
| Agent Through Network | 57% | 52% | n.s. |
| Mortgage through… | |||
| Networks | 27% | 25% | |
| Existing Rltshp | 37% | 22% | |
| Referral | 19% | 12% | |
| Own research | 10% | 22% | |
| Solicitation/Ads | 5% | 14% | |
| Other | 2% | 5% | |
| Total | 100% | 100% | p<.05 |
| Real Estate Agent Characteristics for Buyers Only | |||
| Female Agent | 54% | 51% | n.s. |
| Own Race Agent | 100% | 60% | p<.001 |
| How Used Internet (among those using internet) | |||
| Listings | 84% | 87% | n.s. |
| Mortgages | 36% | 56% | n.s. |
| Info. On Communities | 51% | 44% | n.s. |
Column 3 reports p-values for test of difference between blacks and whites taking into account sample design. Sample size is 418 except for items on real estate agents (n=169); mortgage questions (n=238); and internet questions (n=93); and mortgage and internet questions, n=60).
However, several of these racial differences could be a function of the lower likelihood that African Americans are searching to rent as opposed to buy since in our Detroit data, 16 percent of whites compared to 41 percent of African Americans were looking to rent. Table 3 explores this possibility and also whether any observed racial differences might be explained by social and demographic differences. For each search strategy showing a racial difference in Table 2, I test three logistic regression models: (1) Model 1 includes race of respondent only; (2) Model 2 adds whether the search was to rent or to buy; and (3) Model 3 adds age, gender, education, income, presence of children in the home, and number of years living in the metropolitan area.
Table 3.
Logistic regression models for use of various search strategies, odds ratios, 2004 Detroit Area Study.
| Use of each search strategy in recent housing searches | |||||
|---|---|---|---|---|---|
| Open House | Real Estate Agent | Community Organization | Property Management Firm/Rental Agent | Internet | |
| Model 1 | |||||
| African American Respondent | .42* | .61+ | 5.69* | 4.87** | .42** |
| Model 2 | |||||
| African American Respondent | .59 | .96 | 4.67* | 3.70* | .32** |
| Homebuyer | 8.74*** | 8.61*** | .48 | .34* | .43* |
| Model 3 | |||||
| African American Respondent | .67 | .96 | 3.32+ | 3.60* | .32*** |
| Homebuyer | 7.96*** | 8.08*** | .85 | .52 | .37** |
| Age | .99 | 1.01 | 1.00 | 1.00 | .96+ |
| Female | 1.03 | 1.10 | 2.91 | 2.66 | .78 |
| Less HS | .76 | .55 | 10.98** | 1.00 | .28 |
| HS Degree | .40* | .47+ | 1.05 | 1.73 | .45 |
| Some College | .46+ | .55 | 1.57 | 2.31 | .21** |
| Children in HH | .82 | 2.26* | .84 | .53 | 1.44 |
| Time in Metro Area | 1.03* | .99 | .97 | 1.00 | .98 |
| Yearly Family Income <20k | .56 | .86 | 7.81* | 10.75* | .76 |
| $20k to <$40k | 1.21 | .57 | 21.14*** | 5.35+ | 1.08 |
| $40k to <$80k | 1.48 | .76 | n/a | 12.75* | .96 |
| (n) | (414) | (414) | (414) | (414) | (414) |
p<.10;
p<.05;
p<.01;
p<.001
Omitted categories are: College Degree; Family Income > $80,000/year except for Community Organization analysis where omitted category is $40,000 and above because of no cases where > $80,000/year used Community Organizations.
Once renter/buyer status is held constant, blacks and whites are equally likely to use real estate agents and to attend open houses, the two housing search strategies most used by those looking to buy a home. This pattern persists in Model 3, where additional controls are included. However, further analyses exploring the use of open houses uncovered an interesting interaction effect. Specifically, although the multivariate models reported in Table 3 show that the direct effect of race disappears after controlling for buyer/renter status, including an interaction between race and buyer/renter status reveals a more complex relationship. Among renters, African Americans are significantly (p<.01) more likely than whites to attend open houses (12 percent versus 1 percent); but among buyers, whites are significantly (p<.05) more likely than African Americans to attend open houses (41 percent versus 23 percent). This interaction effect is confirmed in a logistic regression model (results not shown) where the interaction between race and buyer/renter status is statistically significant (p<.01).
For the two strategies used by renters more than buyers (property management firms/rental agents and community organizations), even after controlling for buyer/renter status (Model 2), and background characteristics (Model 3), African Americans are still more likely than whites to use these strategies. To be sure, social class matters. For example, the odds that individuals with incomes less than $20,000 and between $20,000 and $40,000 will use community organizations are 8 and 21 times larger, respectively, than those whose income exceeds $40,000.10 However, these background controls reduce but do not eliminate the racial effects: the odds that African Americans will use either of these resources are just over three times larger than those of whites.
Although there was no difference in Table 2 between blacks and whites in the use of newspapers, further analyses uncovered another interaction. Specifically, black renters are significantly (p<.01) less likely to consult newspapers in their search for an apartment than white renters (31 percent and 53 percent, respectively). Black and white buyers, however, use newspapers at the same rates (60 percent and 66 percent, respectively), a difference that is not statistically significant. Our multivariate analyses (results not shown) confirm the statistical significance of this interaction (p<.01) even after including other control variables.
The final column in Table 3 shows that even after holding constant social, economic, and demographic characteristics, being African American decreases the odds of using the internet by a factor of .32. Indeed, comparing the coefficients across the three models, the racial digital divide increases after taking into account background characteristics.
Returning to Table 2, there are no statistically significant racial differences in terms of which strategy searchers used “the most”: whites and blacks alike used their “own research” the most, and were equally likely to say they used “professionals” the most. Although overall racial differences are not statistically significant, there is again an interesting interaction: while 36 percent of African American renters say that the most used strategy was their networks, just 9 percent of white renters relied on networks the most. White renters, by contrast, are more likely than black renters to use most often their “own research” (61 percent vs. 35 percent). At the bivariate level, these differences are of borderline statistical significance (p=.07), but a multinomial logistic regression model (not shown) reveals a statistically significant interaction effect (p<.05). That is, the odds are greater that black renters over white renters will report using their networks as compared their “own research” the most.11 The greater use of networks by black renters is consistent with Farley’s (1996) findings in Detroit in 1992.
Thus far, we see that when demographic characteristics and whether the search is to rent or to buy are held constant, blacks and whites—especially buyers—are quite similar in the strategies they use to find housing. In addition to reflecting the differences in the practicalities of a search to rent versus buy, this owner/renter search type variable is also to some degree a proxy for social class. Thus, we can conclude that social class shapes the strategies used by searchers. But, there are some interesting persistent racial differences. First, similar to previous research (Newburger 1995), African American buyers are less likely than white buyers to attend open houses. And there is evidence that African American renters both are more likely than whites to seek assistance from organizations—although the level of use of these methods is quite small—and also to be more likely to use informal (networking) strategies.
However, moving beyond the type of strategy people use, to more details about the features of these information sources and how people find them, reveals even greater racial differences (see Table 2). First, African Americans were less likely than whites to get their mortgage through an existing relationship with a company or bank and more likely to either use their “own research” (looking for ads) or be directly solicited by a broker or bank. Controlling for background characteristics in a multinomial logistic regression model does not reduce the racial differences (results not shown): the odds of finding a mortgage through their own research or advertising, as compared to an existing relationship with a bank. are 7.6 (p<.01) and 9.8 (p<.01) times greater for African Americans than for whites.
Second, whereas Detroit area whites and blacks are equally likely (57 percent and 52 percent, respectively) to find their real estate agent through the use of their networks, and equally likely to have a female agent, there is a quite dramatic difference in the race of their real estate agent: nearly all white Detroit homeseekers had a white real estate agent and fully 6 out of 10 African American Detroit homeseekers had an African American real estate agent. So, while blacks and whites both seek the help of real estate agents, there is a great deal of segregation in terms of who those agents are.
Third, while there were racial differences in the likelihood of using the internet, among those who used it, the purposes were similar to some degree: blacks and whites were equally likely to search for listings and for information on the communities they were considering,12 but when asked “Did you look for information on the Internet about financing your home or getting a mortgage,” 56 percent of African Americans said “yes,” compared to just 34 percent of whites. Though a substantial difference, the relationship is not statistically significant, due in part to small sample sizes.
Research Question #2: Are There Racial Differences in Experiences Searching for Housing?
Existing research on housing searches and audit studies of housing discrimination suggest that African Americans would look at fewer homes, have a longer search time, and experience greater levels of difficulty when searching for housing than whites. In this analysis, African Americans’ reports of explicit housing discrimination is not assessed; rather, I analyze racial differences in reports about the ease of the search and of finding a mortgage, how long the search took, the number of homes inspected, and how many offers to buy or applications to rent the individual submitted during their most recent search. Examining the effect of race, net of social class, will answer questions about whether social class explains these differences. Any persistent racial differences may be a consequence—though there is no direct way to test this with these data—of the kinds of disparate treatment identified in audit studies.
As shown in Table 4, whites and African Americans spend about equal amounts of time on their housing searches. About one-half of each group spends less than two months, and the other half more than two months. Consistent with Newburger’s (1995) study, blacks inspect on average fewer units than whites; but the difference is not statistically significant. Racial differences emerge in the number of offers tendered (by buyers) and applications submitted (by renters). Twice as many blacks as whites submitted applications for three or more apartments; and three times as many blacks as whites tendered three or more offers before they successfully purchased a home. The former is statistically significant, and the latter is borderline statistically significant (p<.10).
Table 4.
Housing search experiences by race of respondent, 2004 Detroit Area Study.
| Search Experiences | Whites | African Americans | p-value* |
|---|---|---|---|
| Length of the Search | |||
| <1 month | 26% | 27% | |
| 1–2 months | 28 | 24 | |
| > 2 months | 46 | 48 | n.s. (n=416) |
| Total | 100% | 100% | |
| Mean number places inspected | 9.3 | 7.1 | |
| Std. Error | (1.50) | (0.83) | n.s. (n=414) |
| # of Apt. Applications Submitted | |||
| 0–1 | 77% | 41% | |
| 2 | 10 | 29 | |
| 3 or more | 14 | 30 | p<.01 (n=155) |
| Total | 100% | 100% | |
| # of Offers Tendered | |||
| 0–1 | 79% | 69% | |
| 2 | 16 | 16 | |
| 3 or more | 5 | 15 | p<.10 (n=271) |
| Total | 100% | 100% | |
| Difficulty of Search | |||
| Extremely/Very Difficult | 13% | 27% | |
| Somewhat Difficult | 31 | 25 | |
| Only a little/Not at all difficult | 56 | 48 | p<.10 (n=416) |
| Total | 100% | 100% | |
| Difficulty Getting Mortgage | |||
| Very Easy | 57% | 42% | |
| Somewhat easy | 25 | 33 | |
| Neither easy/diff. | 10 | 15 | |
| Somewhat difficult | 4 | 7 | |
| Very difficult | 5 | 4 | n.s. (n=236) |
| Total | 100% | 100% | |
| Felt taken advantage of | 8% | 21% | p<.05 (n=415) |
Column 3 reports p-values for test of difference between blacks and whites taking into account sample design.
The number of offers/applications it takes before one is successful at securing an apartment or home may be influenced by income, since lower income may result in a landlord’s unwillingness to rent or may lead to a lower offer than possible competitors in the purchase of a home. However, as Model 3 in Table 5 shows (columns 1 and 2), racial differences persist after controls are introduced: African American renters submit 1.7 more applications than their white counterparts, and African American buyers tender about 0.5 more offers then whites when purchasing a home.
Table 5.
Multivariate models of housing search experiences, logistic (odds ratios) and linear regression, 2004 Detroit Area Study.
| Number applications | Number offers | Difficulty of Housing Searcha | Taken Advantage of | |
|---|---|---|---|---|
| (linear) | (linear) | (linear) | (logistic) | |
| Model 1 | ||||
| African American Respondent | 2.19** | .49+ | .32+ | 3.06* |
| Constant | −1.02 | .54+ | 1.95*** | n/a |
| R-squared | .11 | .03 | .01 | n/a |
| Model 2 | ||||
| African American Respondent | n/a | n/a | .39* | 3.76* |
| Homebuyer | n/a | n/a | .22 | 2.31+ |
| Constant | n/a | n/a | 1.69*** | n/a |
| R-squared | n/a | n/a | .02 | n/a |
| Model 3 | ||||
| African American Respondent | 1.66** | .46* | .31* | 4.4*** |
| Homebuyer | n/a | n/a | .24 | 2.6+ |
| Age | .002 | .01 | .006 | .99 |
| Female | −1.02 | −.07 | .34* | .63 |
| Less HS | .65 | .13 | −.08 | .39+ |
| HS Degree | −1.28+ | −.32+ | −.16 | .42 |
| Some College | .71 | −.22 | −.12 | .49 |
| Children in HH | .62 | .40** | .10 | .80 |
| Time in Metro Area | −.01 | −.01 | −.01* | .99 |
| Annual Family Income <$20k | 2.00* | .11 | −.18 | 1.45 |
| $20k to <$40k | .84 | .40+ | −.19 | 2.29 |
| $40k to <$80k | −.18 | .03 | −.40* | 1.66 |
| Constant | −.02 | .35 | 1.92*** | n/a |
| R-squared | .25 | .10 | .08 | n/a |
| Sample Size | (155) | (270) | (414) | (414) |
p<.10;
p<.05;
p<.01;
p<.001
Omitted categories are college degree and yearly family income greater than $80,000.
Difficulty of search is a five-point scale, reverse coded so that a 5 is “extremely difficult”.
Returning to Table 4, on the subjective question of how easy or difficult the respondent felt their housing search was, overall, the percentage of blacks who said it was “extremely” or “very” difficult is double that for whites. Table 5, column 3, shows that when background controls are included, a borderline statistically significant racial difference (p<.10) remains. A second subjective question about the housing search shows a more striking racial difference. Respondents were asked a simple yes/no question: “During the search, was there ever an occasion when you thought you were being taken advantage of?” As reported in Table 4, African American searchers were significantly more likely than whites to say “yes” to this question: one in five African Americans thought they had been taken advantage of, compared to fewer than one in ten whites. As shown in Table 5, after social and demographic controls are introduced, the racial difference persists, and in fact increases: the odds that an African American searcher will say they were taken advantage of during their most recent search are over four times greater than for a white searcher.
Research Question #3: Are There Racial Differences in the Racial Characteristics of Search Locations?
To answer this question, I constructed several variables to look both individually and collectively at the racial characteristics (based on 2000 Census data) of the communities where people said they had searched for a house or apartment in the past 10 years. The 33 communities labeled on the map were assigned to one of seven mutually exclusive categories, defined as follows (shown in parentheses is the number of communities in each category):
All white (n=19): communities where 85% or more of its residents are white, and there are not more than 10% of any other single racial/ethnic group;
Mostly white (n=4): communities where 70% or more of the residents are white and there are fewer than 25% of any single other racial/ethnic group;
All black (n=2): communities where 85% or more of its residents are black, and there are not more than 10% of any single other racial/ethnic group;
Mostly black (n=2): communities where 70% or more of the residents are black and there are fewer than 25% of any single other racial/ethnic group;
Black/White Mixture with White Majority (n=2): communities where 50%–69% of the residents are white and 11%–44% are black.
Black/White Mixture with Black Majority (n=2): communities where 50%–69% of the residents are African American and 11%–44% are white;
Three-Group Mixture (n=2): communities where three racial/ethnic groups have populations in excess of 10%.
The first question is: What percentage of white and black respondents included among their search locations a community of a particular type? Table 6 shows a clear, if unsurprising pattern for whites. Slightly more than seven out of ten white Detroiters searched in the last 10 years in a community that was “all white,” 40 percent in a community that was “mostly white,” and 20 percent searched in none of the communities on the map. As noted earlier, the latter group is likely to include many additional whites who searched in “all white” or “mostly white” communities, given that the majority of the CDPs not labeled on the map are all or mostly white. For the remaining community types, the percentage of whites who had searched in the areas dropped to less than 10 percent.
Table 6.
Racial characteristics of housing search locations by respondent race, 2004 Detroit Area Study.
| Whites | African Americans | Race Alone | Race w/Controls | ||||
|---|---|---|---|---|---|---|---|
| (odds ratio) | (odds ratio) | ||||||
| Mean number of selected communities on map | 2.84 | 3.40 | n.s. | ||||
| Std. Error (n=382) | (.303) | (.289) | |||||
| Kinds of communities selected on map | |||||||
| All white | 73% | 47% | ** | .341 | ** | .312 | ** |
| All black | 4% | 44% | *** | 18.346 | *** | 17.806 | *** |
| Mostly white | 40% | 30% | n.s. | .663 | ns | .847 | ns |
| Mostly black | 3% | 48% | *** | 31.295 | *** | 28.728 | *** |
| Mixed Black-White | 2% | 6% | n.s. | 2.592 | ns | 1.204 | ns |
| White Majority | |||||||
| Mixed Black-White | 9% | 40% | *** | 6.815 | *** | 9.474 | *** |
| Black Majority | |||||||
| Mixed Three Groups | 5% | 19% | ** | 4.067 | ** | 5.248 | *** |
| None on the map | 20% | 5% | * | .216 | * | .213 | * |
| (n) | (382) | (379) | (379) | ||||
Notes: p<.05;
p<.01;
p<.001
This analysis is based on the sub-sample of all active searchers who completed the map task. Compared to earlier tables, in this analysis we lose 32 respondents who for one reason of another indicated on the map that they had not searched in the last 10 years, and yet reported they had earlier in the survey. Regression analyses are based on slightly fewer cases due to missing data on predictor variables. Control variables in the regressions are: type of search (own versus buy), age, sex, education, presence of children, number of years in metropolitan area, and income.
African American patterns of search locations are far more diffuse than whites’. There was no single neighborhood “type” that the majority of blacks searched in. Rather, 40 to 48 percent of African Americans (who had searched at all), had in the last 10 years searched in all white, all black, mostly black, and mixed black-white with black majority communities. Very few did not search in any of the communities labeled on the map (5 percent), and as was the case for whites, very few searched in the mixed white-black with white majority communities.13 Interestingly, about 1 in 5 African Americans had searched in the three-group mixture communities—compared to just 5 percent of whites. Columns 3 and 4 in Table 6 test the effect of an individual’s race on whether or not they had searched in the last ten years in each of the different community types before and after controlling for social and demographic characteristics. Not surprisingly, the racial differences persist and are in some cases quite dramatic. After controlling for background characteristics, the odds are 9 times greater that a black searcher will have searched in a mixed black-white, black majority community than a white searcher.
Respondents rarely look in only one location for a place to live. Indeed, on average, as reported in Table 6, respondents searched in about 3 different labeled locations. This raises the question: Did respondents search in a variety of community types, or were they restricted to just one kind? To answer this question, our second analysis paints a portrait of the combination of locations in which respondents searched. To capture this dynamic, a four-category variable was constructed: (1) Own Majority Only includes those individuals who searched in the last 10 years only in communities where their racial group was in the majority (50% or more); (2) Other Majority Only is comprised of those who searched in the last 10 years only in communities where the other racial group was in the majority (50 percent or more); (3) Both Own Majority and Other Majority consists of those whose search locations in the last 10 years included both communities where their own group was in the majority and communities in which the other group was in the majority; (4) None on Map are those who did not search in any of the labeled communities.
Panel A in Table 7 shows the distribution of this variable separately for blacks and whites. Just about two out of every three whites in Detroit searched exclusively in communities where their own group was in the majority. By contrast just 32 percent of blacks had searched only in communities where at least 50 percent of its residents were African American. For African Americans, the modal category is “Both Own and Other”—that is, just over one-half of African Americans have searched in both a majority black and a majority white community in the last ten years; just 14 percent of whites have done the same. Finally, only a very tiny number of whites (1 in 100), but just over 1 in 10 African Americans searched only in communities where their own racial group was fewer than 50 percent.
Table 7.
Racial Characteristics of mixture of housing search locations, 2004 Detroit Area Study.
| Own Only | Other Only | Both Own and Other | None on Map | Total | |
|---|---|---|---|---|---|
| PANEL A | |||||
| Bivariate Analysis | |||||
| White Searchers | 64%*** | 1 | 14 | 20 | 100% |
| African American | 32% | 11 | 52 | 5 | 100% |
| Searchers (n) | (382) | ||||
| PANEL B | |||||
| Multinomial Logistic Regression Results | |||||
| Reference Category | (Relative Risk Ratios) | (Relative Risk Ratios) | (Relative Risk Ratios) | ||
| Model 1 | |||||
| African American (n=382) | -- | 16.76*** | 7.40*** | .519 | |
| Model 2 | |||||
| African American | -- | 18.32*** | 8.32*** | .50 | |
| Homebuyer | -- | .16* | .47 | .60 | |
| Age (in years) | -- | 1.00 | .95* | .99 | |
| Female | -- | 1.39 | .72 | 1.25 | |
| Less HS | -- | .09** | .29* | .34 | |
| HS Degree | -- | .08** | .47 | 2.39 | |
| Some College | -- | .04** | .34 | 1.06 | |
| Children in HH | -- | 2.02 | .55 | .37* | |
| Time in Metro Area | -- | .99 | 1.00 | .98 | |
| Yr. Fam. Income <$20K | -- | 1.61 | 1.74 | .28 | |
| $20K to <$40K | -- | 6.02* | 6.00* | 1.61 | |
| $40K to <$80K (n=379) | -- | .70 | .75 | 1.47 | |
Note: p<.05;
p<.01;
p<.001
Using this four category variable as a dependent variable in a multinomial logistic regression model reveals that social class and other background characteristics do not diminish these racial differences. With the “Own Majority Only” category as the reference, the third column in panel B reports the relative risk ratios, and reveals that in model 2, which includes social class and background characteristics, there is no attenuation of the effect of respondent race. In fact, if anything, the effect of race becomes greater. The odds that one has searched in both types of communities, relative to searching only in a community in which one’s own group is in the majority, are, for example eight times greater for African Americans than whites. By contrast, the influences of education and income—two indicators of social class—are smaller.14
DISCUSSION
While primary attention in research into the causes of racial residential segregation has focused on three key factors of money, preferences, and discrimination, less attention has focused on the mechanisms through which these factors operate. In this paper, I suggest that through a better understanding of how people actually search for housing, and what they experience when they do, we can ask better questions about where in the process—and what elements of that process—contribute to moves that perpetuate residential segregation.
Our analysis of Detroit area residents who have searched for housing in the past 10 years finds that whites and African Americans do not differ overwhelmingly in their use of various housing search strategies, especially after controlling for one loose proxy of social class—whether the search is to rent or to buy. To the same degree, whites and blacks—once controls are included—generally invoke networks, “leg work,” and professional assistance when trying to find a place to live. Some racial differences persist, including one that may become increasingly important: the use of the internet. Given the rapid growth of the internet in renting and selling housing, the observed racial digital divide is a point of some concern. It is likely that the kinds of homes and apartments marketed on the internet differ from those available through other low-tech means and so those who do not use this medium may be at a disadvantage. If the move toward internet use in the marketing of housing increases—as is likely—then this disparity will only increase.
Despite these general similarities between blacks and whites, this analysis suggests that scratching the surface of search strategies reveals in some cases substantial racial differences with potentially important consequences. For example, there are differences in how blacks and whites find mortgages. Detroit area whites find them through previous relationships with banks or other institutions; African Americans in the Detroit area—more so than whites—find their mortgage by responding to direct mail advertising or hearing about them through other advertising sources. One could speculate that drawing on existing relationships will lead to a more positive experience than responding to traditional advertising campaigns, some of which may be predatory lenders with aggressive marketing strategies and/or inferior products. Our data cannot assess the quality of mortgages obtained through these different means, but these observed racial differences call for further investigation into this possibility.
An even more striking racial difference arises when looking at who the real estate agents are that white and black homebuyers use. That is, while the techniques for searching for a house to buy do not generally differ greatly for black and white buyers, it is clear that searches are racially segregated in terms of who is assisting the buyer. Whites use white real estate agents; African Americans to a great extent use African American real estate agents. To the degree that real estate agents are marketing homes in neighborhoods that are themselves racially segregated, then this race matching of real estate agent and client may be a nexus at which segregation is perpetuated. To be sure, any segregation of agents and housing markets is itself likely a function of existing patterns of segregation and discrimination.
With respect to self-reported housing search experiences, although there are some similarities between whites and blacks, there are troubling signs. Whereas blacks and whites in Detroit report equally lengthy housing searches, and inspect about the same number of units, other measures suggest that African Americans face more hassles and have to put more effort into securing a place to live. African American renters and buyers submit more applications and tender more offers than did their white counterparts. In addition, blacks report significantly more difficult housing searches than whites and are much more likely to feel that they were “taken advantage of” during their search. Both of these patterns point to the greater financial and psychic costs of housing searches for African Americans than whites, a finding consistent with the implications of experimental housing discrimination audit studies (Yinger 1995; Turner et al. 2002). These differences in disadvantage and hassles cannot be explained solely by economic status; even after controlling for a variety of background characteristics the racial differences in many cases persist.
Although we know from other research that whites are quite likely to move into an all-white neighborhood when they do move (e.g., South and Crowder 1998; Quillian 2002), it is of interest to know the range of different communities in which whites actually search. In particular, do whites and blacks take even that first step towards making an integrative move by searching for housing in a racially mixed community or in a community that has a predominance of a racial group that is not their own? Insights into this question, to date, have come from data that is only indirect, and largely based on reports of hypothetical neighborhood preferences. The present study provides a rare look at the range—or lack thereof—of the racial compositions of places where individuals have actually searched for housing. Interestingly, the results are quite consistent with the patterns reported in hypothetical neighborhood preferences surveys.
First, white Detroiters’ search locations are highly segregated. The vast majority of whites (64 percent) searched in the last 10 years only in communities where they were in the majority, a figure that is likely a conservative estimate since another 20 percent searched in none of the communities on the map (the vast majority of which are likely to also fall in the “majority white only” category). The pattern for African American respondents contradicts the proposition that African Americans’ preferences for “all black” neighborhoods—or even for racially mixed neighborhoods—causes segregation (Thernstrom and Thernstrom 1997; Patterson 1997). That is, our portrait reveals that the racial characteristics of the communities where African Americans search for housing in the Detroit metropolitan area is one of great diversity. Combining those who searched in both majority white and majority black communities with those who searched exclusively in majority white communities, accounts for 63 percent of all African American Detroiters who had searched in the past 10 years. Controlling for an individual’s social class characteristics does not eliminate the racial differences in search locations.
Several limitations of the findings just reported are worth highlighting. First, the study was limited to a single metropolitan area. Cities with less segregated histories or different demographic patterns may show quite different results, particularly in terms of the racialized nature of housing search locations. However, given the topic of study—particularly with respect to housing search locations—it would be difficult to conduct a similar study at a national scope, since housing markets are local phenomenon. The selection of Detroit as the study site may also be problematic since it is the most segregated metropolitan area in America and further, it has had among the smallest declines in segregation over the past several decades (Lewis Mumford Center, 2001). In addition, although rapidly growing, the Asian and Latino populations in the Detroit metropolitan area are still quite small (about 3% of each group). “Chocolate City, Vanilla Suburbs” (Malbix/Mix Records, BMI 1976) still aptly describes the Detroit metropolitan area.
However, despite the fact that the results can only be generalized to the Detroit metropolitan area, there are other cities that have much in common with Detroit’s history and patterns of segregation, including places like Cleveland, Milwaukee, Rochester, Newark, St. Louis, and other older cities in the Rust Belt. In addition, it is worth noting that in the early 1990s, Detroit was one of four cities in the Multi-City Study of Urban Inequality (MCSUI), a study that included measures of racial residential preferences. In analyses of the MCSUI data, Krysan (2002a) and Farley et al. (1997) found that although Detroit was often more extreme in its attitudes than the other cities, the general patterns were similar to those in Atlanta, Boston, and Los Angeles.
More specific limitations relate to the findings on the search locations themselves. First, although the maps allowed us to gather information on a much larger number of communities than prior studies, there were many communities that were left off the map. Indeed, 20 percent of whites who did search in the past 10 years had not searched in any of the communities that were labeled. This means that we know nothing about the neighborhood racial compositions where these individual searched—although given that most of the omitted communities were overwhelmingly white, the probabilities are high that their searches took them to segregated white neighborhoods.
A second limitation is more subtle, but suggests a further caveat, especially with respect to the conclusions about whites’ search locations. Several of the majority black communities labeled on this map had quite low median housing values. This is in part because all of the “all” black communities on the map are located in the city of Detroit, where housing values are compressed. In essence, the range of housing values open to whites who might consider a majority black or mixed neighborhood is limited. The one clear exception is the inner-ring suburb of Southfield, which is a prosperous upper middle class community where housing values and median family income are quite high and where the population is 54 percent African American. It is noteworthy that just 8 percent of whites had searched in this community in the last 10 years. Nevertheless, the caveat holds that in the Detroit metropolitan area, there are simply fewer—in terms of numbers—different communities across a range of housing values that would constitute a “diverse” move for whites (one where they moved to a community that was anything but all or mostly white). The results for whites should be viewed in this context.
To some degree, then, it is the portrait of housing search locations for African Americans that is particularly compelling—and largely for what it does not show. The results do not show that blacks search exclusively in all-black communities. Nor even that their searches are restricted to “mostly” black communities. To be sure, African Americans do search in such communities; but the majority of African Americans also consider communities with a substantial white population. Indeed, 11 percent of African Americans have searched only in communities where whites predominate; and the majority (52 percent) searched in both majority and minority black communities. Thus African American preferences, as expressed through search locations, are more, rather than less, supportive of integration.
Discrimination in the form of hassles and barriers also appear to be greater for blacks than whites—and these racial differences hold even after controlling for the effect of social class. Although housing discrimination is not directly tested in this analysis, the apparent pattern that many African Americans search in racially mixed and also predominately white neighborhoods, but many presumably do not end up moving there, opens up further questions about what happens between the search and the move that leads each group into largely homogeneous communities.
CONCLUSION
As is true of most exploratory studies, many more questions are raised than are answered by this analysis. For example, one theme surfaced at several points in this analysis, is rarely considered explicitly in discussions of the causes of segregation, and is in need of further exploration to determine its possible implications: the role of networks. Whites and blacks both use networks when looking for housing but black and white renters differed in their level of use: 58 percent of black renters talked to friends and/or relatives, but just 36 percent of white renters did (p<.05). And African American renters more so than white renters reported having used networks the most (36 percent of African American renters relied on this strategy the most versus 9 percent of white renters). Given what we know from the social networks literature about homophily (McPherson et al. 2001), the person giving the assistance and receiving the assistance are very likely to be of the same race. This further suggests that, given the high levels of existing residential segregation, the information provided by these segregated networks may result in segregated moves. Future studies should examine the degree to which the segregated networks within which renters operate result in segregated housing information that, in turn, leads white renters to live in whiter neighborhoods and black renters to live in blacker neighborhoods.
The role of networks among homebuyers emerged less directly but again raises a series of further questions. Although white and black homebuyers relied most (and equally) on professional resources such as real estate agents, they were clearly not using the same real estate agents: white Detroiters almost universally consulted with white real estate agents; African Americans, in great numbers (60 percent), consulted with African American real estate agents. And to a great extent, it was through their social networks that they got their agents: over one-half of both black and white buyers said they were referred to their agent by someone in their network.15
As an aside, these results suggest that the majority of those in the market to buy a home—white and black—do not get their real estate agents in the way that housing audits test. In the manner of a housing audit, a homeseeker sees an advertisement in the newspaper and calls on the real estate agent who was listing that home. The audit study then proceeds to compare how this real estate agent/agency, which was the result of a “cold call,” treats the white and black client. But for at least one-half of the Detroiters in our study, this is not how they approached the task. Rather, they sought the advice and suggestions of friends, relatives, neighbors, and other acquaintances when trying to find an agent with whom to work. Not only is the agent choice not “random,” but there is also the strong possibility that this personal connection will lead to different—and presumably better—service than a “cold call.”16
The quite profound race-matching between real estate agent and client—likely due in part to the role of networks in making these matches—raises the possibility of another way in which the real estate industry may contribute to the perpetuation of housing segregation. That is, audit studies have focused on—and found evidence of—disparate treatment by realtors who steer whites and blacks to different areas or who are less helpful to black clients than to white clients. But missing is the possibility of an equally, if not more pernicious and intractable consequence arising from the substantial race-matching between agent and client that happens in housing searches. Several questions merit additional attention. First, what are the consequences of this race-matching? To what extent do African Americans who have white real estate agents end up in more or less segregated neighborhoods than their counterparts whose agents are African American? One could imagine equally plausible competing hypotheses. On the one hand, African Americans may face less discrimination and steering from their African American agent than they would a white agent—and so may end up in more racially diverse or “whiter” neighborhoods. Or, alternatively, African American agents may specialize in and have more knowledge about real estate in predominately black neighborhoods (due at least in part to discrimination in the real estate industry itself), and so may market more to those areas, resulting in more segregated moves for their clients. This raises a related question: Do black and white agents steer their clients toward certain neighborhoods—either consciously or unconsciously—simply by virtue of the areas they are familiar with and hold expertise about? This in turn begs the question: Do African American and white realtors represent and market very different parts of the housing market? That is, are real estate companies themselves integrated or segregated—both within a particular agency as well as across all of the agencies serving a metropolitan area?
One is then compelled to ask the underlying question: Why do whites and blacks end up with white and black real estate agents, respectively? Is it because people simply feel “more comfortable” with someone of their own race? Or is it because our lives continue to be so segregated that our networks—where we get our real estate agents—are thus also segregated? Racial discrimination may also be implicated: African American clients, concerned about possible discrimination at the hands of white agents, may seek out African American agents. And African American real estate agents, for their part, may face discrimination in their jobs so they end up marketing homes largely in black areas—either by choice or out of necessity. Finally, white clients—living in whiter neighborhoods—may discriminate against African American agents, choosing not to hire them to represent them either as buyers or sellers. Unfortunately, the data at hand do not answer any of these questions, but instead point to an area ripe for future research.
Standard sample surveys, in-depth interviews of housing searchers, and ethnographic studies of the real estate industry and real estate agent-client relationships are called for so that we can better understand the answer to these and other questions about how the broad patterns of racial differences in search strategies, behaviors, and experiences may or may not translate into moves that either perpetuate or reduce segregated housing patterns. Other results reported in this paper suggest additional research questions beyond the issue of networks. For example, this study demonstrated that African Americans search in a variety of kinds of neighborhoods but it is likely that they often end up moving into segregated neighborhoods. First, is this the case? And second, if they search in, but do not end up moving to, more diverse or whiter neighborhoods, why not? What are the barriers they face? Similarly, do those whites who search in more racially diverse areas move into them? And if not, why not? With respect to African Americans, how do experiences with discrimination shape search behaviors and locations? Do African Americans who experience discrimination limit the racial composition of their housing search locations? Conversely, do African Americans who search in whiter areas report more negative experiences and file more applications/submit more offers before they are successful? And, to what extent do search strategies shape the kinds of neighborhoods people search in? For example, does using the internet expand or restrict the kinds of neighborhoods, vis a vis racial composition?
This exploratory study of racial differences in housing searches is therefore suggestive of fertile ground for future studies. Directing our attention to these earlier steps in the housing mobility process helps us break out of the pattern of focusing on money, preferences, and discrimination and highlights the need to consider the mechanisms through which these factors operate so as to better understand how segregation is perpetuated. In doing so, hopefully, we can discover how to break the cycle of housing segregation that continues to characterize many of our nation’s largest cities nearly four decades after the Fair Housing Act was signed into law.
With respect to the finding of low rates of black movement into white neighborhoods, Quillian (2002) notes that although blacks are less likely than whites to move into white neighborhoods, such moves do occur with some regularity—about 10 percent of all blacks move into neighborhoods with fewer than 10 percent black residents.
This is because these questions require a match between search features and search locations. In our survey, respondents were asked about the features of their most recent search (length of time, difficulty, information sources used, etc.) but they were asked about any locations in which they had searched at any point in the last 10 years. Thus, the latter could cover multiple searches, whereas the former refers only to the most recent search.
The calculation of the response rate follows the formula designated as “Response Rate 2” in the AAPOR Standard Definitions, 2006.
Sixty-six percent of respondents indicated they had moved in the past 10 years. Of these, 10 percent indicated that they were “not involved at all” in the search. Because uninvolved respondents would not be able to answer detailed questions about the housing search with much accuracy, they were skipped out of the housing search questions. Thus, our analysis is limited to the 63 percent of respondents who had both searched for housing in the preceding 10 years and said they were involved “a little” or more in it.
Our mortgage questions were asked of all respondents who were not currently living in rentals and who had obtained a mortgage in the past 5 years. We limited the time frame for this question because of concerns about memory limitations. Individuals may have refinanced their mortgages on their current homes without having moved. Thus, all analyses of mortgages include all individuals—whether they were home searchers or not—who had purchased a mortgage in the previous five years.
In terms of population distribution, the 29 places identified on the map contain 75 percent of the population living in Wayne County (where the city of Detroit is located), 33 percent of the population living in Oakland County, and 46 percent of the population living in Macomb County.
No distinction is made for whether these children are the respondent’s children or simply children in the home.
The imputation model included measures of race/ethnicity, housing tenure, education, median family income of the block group in which the respondent resided, gender, marital status, presence of children under 18, welfare receipt, employment status, and age. The resulting continuous income variable was collapsed into the four categories used in the analysis.
There are no racial differences in the likelihood of having searched but not moved.
Due to a lack of respondents in the $80,000/year or more annual income category who had used community organizations, for this search strategy, the omitted reference category are those who earned $40,000/year or more.
This multinomial logistic regression model includes controls for background characteristics as well as for whether the search resulted in a move or not.
However, restricting the sample to only those who both searched and moved in the prior ten years changed the results. Among this smaller subset of all searchers, there was a racial difference in whether the internet was used to search for information on communities. Among African Americans who used the internet, 82 percent searched for community information; among whites just 45 percent did, a difference that is significant at p<.05 level. The sample size is small (n=52) making it difficult to interpret the finding.
The two communities on our map that met our definition of mixed but white majority have particularly low median family incomes and nearly the lowest median housing values of all those on the map.
Including the number of communities in which a respondent has searched as an additional control variable does not change the substantive findings of the models reported in Table 7.
Interestingly, among African Americans, of those who had an African American agent, 67 percent found them through their networks; just 35 percent of African Americans who had white agents found their agents through their networks (p<.01).
Among Detroiters who found their agent through their networks, 7 percent said they would “not at all recommend” their agent to their friends, compared to 18 percent among those who found their agents in other ways—outside their networks. This difference is borderline statistically significant (p<.06).
An earlier version of this paper was presented at the 2006 Annual Meetings of the Population Association of America in Los Angeles, CA. Reynolds Farley and Mick Couper collaborated with the author on the 2004 Detroit Area Study. They, along with Michael Bader, Kyle Crowder, and Sapna Swaroop, also provided invaluable feedback and suggestions on earlier versions of this paper. Additionally, Michael Bader designed and created the maps that were used to collect the data on housing search locations. Support for the data collection came from the National Science Foundation (SES-0317740), University of Michigan, Ford Foundation, and the University of Illinois at Chicago. Data analysis was supported by Grant Number (1R03HD051677-01A1) from the National Institute of Child Health and Human Development. The articles contents are solely the responsibility of the author and do not necessarily represent the official views of the National Institute of Child Health and Human Development.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
REFERENCES CITED
- Alba Richard D, Logan John R, Stults Brian. How Segregated are Middle Class African Americans? Social Problems. 2000;47(4):543–558. [Google Scholar]
- Alba Richard D, Logan John R. Minority Proximity to Whites in Suburbs: An Individual-Level Analysis of Segregation. American Journal of Sociology. 1993;98:1388–1427. [Google Scholar]
- Charles Camille Zubrinsky. Won’t You Be My Neighbor? Race, Residence and Intergroup Relations in Los Angeles. New York: Russell Sage Foundation; 2006. [Google Scholar]
- Charles Camille Zubrinsky. Neighborhood Racial-Composition Preferences: Evidence from a MultiEthnic Metropolis. Social Problems. 2000a;47(3):370–407. [Google Scholar]
- Charles Camille Zubrinksy. Residential Segregation in Los Angeles. In: Bobo L, et al., editors. Prismatic Metropolis: Inequality in Los Angeles. New York: Russell Sage Foundation; 2000b. pp. 167–219. [Google Scholar]
- Clark WAV, Flowerdew Robin. A Review of Search Models and Their Application to Search in the Housing Market. In: Clark WAV, editor. Modeling Housing Market Search. New York: St Martin’s Press; 1982. [Google Scholar]
- Cronin Francis J. Low Income Household’s Search for Housing: Preliminary Findings on Racial Differences. Washington, DC: Urban Institute; 1979. [Google Scholar]
- Cronin Francis J. Racial Differences in the Search for Housing. In: Clark WAV, editor. Modeling Housing Market Search. New York: St Martin’s Press; 1982. pp. 81–105. [Google Scholar]
- Crowder Kyle. The Racial Context of White Mobility: An Individual Level Assessment of the White Flight Hypothesis. Social Science Research. 2000;29:223–257. [Google Scholar]
- Crowder Kyle. Racial Stratification in the Actuation of Mobility Expectations: Microlevel Impacts of Racially Restrictive Housing Markets. Social Forces. 2001;79(4):1377–1396. [Google Scholar]
- Darden Joe T, Kamel Sameh M. Black Residential Segregation in the City and Suburbs of Detroit: Does Socioeconomic Status Matter? Journal of Urban Affairs. 2000;22(1):1–13. [Google Scholar]
- Farley Reynolds, Schuman Howard, Bianchi Suzanne, Colasanto Diane, Hatchett Shirley. Chocolate City, Vanilla Suburbs: Will the Trend Toward Racially Separate Communities Continue? Social Science Research. 1978;7:319–344. [Google Scholar]
- Farley Reynolds, Steeh Charlotte, Krysan Maria, Reeves Keith, Jackson Tara. Stereotypes and Segregation: Neighborhoods in the Detroit Area. American Journal of Sociology. 1994;100(3):750–78. [Google Scholar]
- Farley Reynolds, Fielding Elaine, Krysan Maria. The Residential Preferences of Blacks and Whites: A Four-Metropolis Analysis. Housing Policy Debate. 1997;8(4):763–800. [Google Scholar]
- Farley Reynolds. Racial Differences in the Search for Housing: Do Whites and Blacks Use the Same Techniques to Find Housing? Housing Policy Debate. 1996;7(2):367–386. [Google Scholar]
- Galster George C. Research on Discrimination in Housing and Mortgage Markets: Assessment and Future Directions. Housing Policy Debate. 1992;3(2):639–683. [Google Scholar]
- Krysan Maria. Privacy and the Expression of White Racial Attitudes: A Comparison Across Three Context. Public Opinion Quarterly. 1998;62(4):506–544. [Google Scholar]
- Krysan Maria. Whites Who Say They’d Flee: Who Are They and Why Would They Leave? Demography. 2002a November;39(4):675–697. doi: 10.1353/dem.2002.0037. [DOI] [PubMed] [Google Scholar]
- Krysan Maria. Community Undesirability in Black and White: Examining Racial Residential Preferences through Community Perceptions. Social Problems. 2002b;49(4):521–543. [Google Scholar]
- Krysan Maria, Couper Mick P. Race in the Live and Virtual Interview: Racial Deference, Social Desirability, and Activation Effects in Attitude Surveys. Social Psychology Quarterly. 2003;66(4):364–383. [Google Scholar]
- Krysan Maria, Farley Reynolds. The Residential Preferences of Blacks: Do They Explain Persistent Segregation? Social Forces. 2002;80(3):937–980. [Google Scholar]
- Lewis Mumford Center. Ethnic Diversity Grows, Neighborhood Integration Lags Behind. Lewis Mumford Center; University at Albany: 2001. [Google Scholar]
- Logan John, Stults Brian, Farley Reynolds. Segregation of Minorities in the Metropolis: Two Decades of Change. Demography. 2002;41(1):1–22. doi: 10.1353/dem.2004.0007. [DOI] [PubMed] [Google Scholar]
- Malbix/Ricks Music, BMI. Chocolate City. 1976. Casablanca Records. [Google Scholar]
- Massey Douglas, Denton Nancy. American Apartheid: Segregation and the Making of the Underclass. Cambridge, MA: Harvard University Press; 1993. [Google Scholar]
- Massey Douglas, Fischer Mary. Does Rising Income Bring Integration? New Results for Blacks, Hispanics, and Asians in 1990. Social Science Research. 1999;28:316–324. [Google Scholar]
- Massey Douglas, Lundy Garvey. Use of Black English and Racial Discrimination in Urban Housing Markets: New Methods and Findings. Urban Affairs Review. 2001;36:470–96. [Google Scholar]
- Massey Douglas, Fischer Mary. The Social Ecology of Racial Discrimination. City and Community. 2004;3:221–43. [Google Scholar]
- McPherson Miller, Smith-Lovin Lynn, Cook James M. Birds of a Feather: Homophily in Social Networks. Annual Review of Sociology. 2001;27:415–44. [Google Scholar]
- Newburger Harriet. Sources of Differences in Information Used by Black and White Housing Seekers: An Exploratory Analysis. Urban Studies. 1995;32(3):445–471. [Google Scholar]
- Patterson Orlando. The ordeal of Integration: Progress and Resentment in America’s “Racial” Crisis. Washington, DC: Civitas/Counterpoint; 1997. [Google Scholar]
- Quillian Lincoln. Why is Black-White Residential Segregation So Persistent?: Evidence on Three Theories from Migration Data. Social Science Research. 2002;31:197–229. [Google Scholar]
- Reskin Barbara. Including Mechanisms in Our Models of Ascriptive Equality. American Sociological Review. 2003;68:1–21. [Google Scholar]
- St John Craig, Clymer Robert. Racial Residential Segregation by Level of Socioeconomic Status. Social Science Quarterly. 1999;81(3):701–715. [Google Scholar]
- South Scott J, Crowder Kyle. Leaving the ‘hood: Residential Mobility Between Black, White, and Integrated Neighborhoods. The American Sociological Review. 1998;73:17–26. [Google Scholar]
- South Scott J, Deane GD. Race and Residential Mobility: Individual Determinants and Structural Constraints. Social Forces. 1993;72:147–167. [Google Scholar]
- Thernstrom Stephan, Thernstrom Abigail. America in Black and White: One Nation, Indivisible. New York: Simon and Schuster; 1997. [Google Scholar]
- Turner Margery, Wienk Ron. The Persistence of Segregation in Urban Areas: Contributing Causes. In: Kingsley G Thomas, Turner Margery Austin., editors. Housing Markets and Residential Mobility. Washington: The Urban Institute Press; 1993. [Google Scholar]
- Turner Margery Austin, Ross Stephen L, Galster George C, Yinger John. Discrimination in Metropolitan Housing Markets: National Results from Phase I HDS 2000. Washington: The Urban Institute, Metropolitan Housing and Communities Policy Center; 2002. [Google Scholar]
- Yinger John. Closed Doors, Opportunities Lost: The Continuing Costs of Housing Discrimination. New York: Russell Sage Foundation; 1995. [Google Scholar]