"Property Values Drop When Blacks Move in, Because...": Racial and Socioeconomic Determinants of Neighborhood Desirability

by David R. Harris
"Property Values Drop When Blacks Move in, Because...": Racial and Socioeconomic Determinants of Neighborhood Desirability
David R. Harris
American Sociological Review
Start Page: 
End Page: 
Select license: 
Select License

David R. Harris

University of Michigan

Are housing prices lower in neighborhoods with high concentrations of black residents? If so, is this relationship evidence of pure discrimination, or can it be explained by considering nonracial neighborhood traits? These ques- tions derive their importance from the link between mobility patterns and residential segregation, and the consequent relationship between high levels of segregation and a host of deleterious outcomes. I assess the magnitude and motivations of racial aversion by conducting a hedonic price analysis of geocoded data from the Panel Study of Income Dynamics. Ifind clear evi- dence of lower property values in neighborhoods with relatively high pro- portions of black residents. However, whether it is blacks' race or their so- cioeconomic status that affects property values depends on whether housing

units are rented or owner-occupied.

''Tt seems like the property values drop

lwhen black families move in," ob- served a white woman when asked why she would move if more than three black fami- lies lived in her neighborhood (Farley et al. 1994:775). This comment resonates with conventional wisdom about the effect of in- tegration on predominantly white neighbor- hoods and is supported by empirical work that finds whites averse to black neighbors. The woman's words reveal a distaste for liv- ing near blacks that is reflected in a low as- sessment of property values. While racial prejudice and discrimination may account for the woman's preference for a segregated neighborhood, there is at least one other pos-

'Direct all correspondence to David R. Harris, Institute for Social Research, University of Michigan, 426 Thompson Street, Ann Arbor, MI 48104-2321 (drharris@umich.edu). Portions of the data were made available by the Inter-univer- sity Consortium for Political and Social Research at the University of Michigan. I have benefited greatly from comments by Christopher Jencks, Greg Duncan, Art Stinchcombe, and anonymous ASR reviewers. Early versions of this paper were presented at the 1997 annual meeting of the American Sociological Association in Toronto, and at the 1997 Institute for Research on Poverty Summer Research Workshop in Madison, Wis- consin.

sible explanation. This same woman went on to say that the reason property values drop when blacks move in is "because they [blacks] do not keep up their houses" (p. 775). This suggests an aversion to neighbors who do not maintain their homes, rather than to blacks per se. Is this woman offering an aversion to rundown property to cover up her racist intentions, or would she actually move if new neighbors of any race did not main- tain their homes?

There are many reasons why people might be averse to black neighbors. One is pure dis- crimination-that whites dislike blacks be- cause they are black (Bobo and Zubrinsky 1996). A second explanation stresses the proxy component of racial aversion-white people avoid black neighbors not because they are black, but because of "other neigh- borhood concerns, correlated with racial mix, which affect people in such a way as to pro- duce cumulative patterns of invasion and suc- cession" (Taub, Taylor, and Dunham 1984: 177). For example, if the only factor in as- sessments of neighborhood quality is the pov- erty rate, highly integrated neighborhoods will be considered less desirable because higher poverty rates among blacks mean that the average mixed-race neighborhood contains more poor people than does the average homogeneous, white neighborhood.

American Sociological Review, 1999, Vol. 64 (June:461-479)

The distinction between pure discrimina- tion and racial proxies is important in part because the two explanations have distinct implications for integration policy. If whites avoid blacks because they are black, then stable integration is unlikely; no matter what policy is pursued, whites will still object to living near blacks. Alternatively, if whites avoid blacks because of characteristics asso- ciated with being black, then stable integra- tion can be achieved through policies that promote racial integration while minimizing undesirable nonracial characteristics.'

In light of this distinction between racial proxies and pure discrimination, my analysis pursues two goals. First, I assess the magni- tude of whites' aversion to black neighbors. Second, I consider the impact of several non- racial neighborhood factors on aversion to black neighbors. This exercise tests the pure discrimination and racial proxy hypotheses. To achieve these goals I employ hedonic price analysis, a method that is common in economics but seldom used in sociology.

One sociological approach to neighborhood desirability stresses the importance of racial prejudice and discrimination in assessments of neighborhood quality (Bobo 1997; Bobo, Klugel, and Smith 1997; Bobo and Zubrin- sky 1996; Zubrinsky and Bobo 1996). To- gether with several collaborators, Bobo ar- gues that negative conceptions of other racial groups, in particular blacks, guide whites' ranking of neighborhoods. As a result, neigh- borhood preferences are directly related to the proportion of residents who are black. Bobo identifies this aversion to black neigh- bors as part of a new "laissez-faire racism."

If Bobo is correct about whites' prefer- ences, then large effects of racial animus on behavioral indicators of neighborhood desir- ability (e.g., moving, property values) might be observed even if all whites do not exhibit extreme prejudice. This potential was real-

' Throughout "nonracial" refers to ecological factors that, while perhaps correlated with racial composition because of historical and contempo- rary discrimination, are not themselves indicators of racial composition.

ized by Schelling (1971), whose tipping model starts with the assumption that whites have varying preferences for black neigh- bors. When a black family moves into a neighborhood, it disturbs the racial equilib- rium. In response, whites who are already in the most integrated neighborhood they will tolerate decide to move. These openings cre- ate opportunities for more black families to join the neighborhood. When the second wave of blacks arrives, their presence prompts the departure of whites who are slightly more tolerant than those who ini- tially fled. Again, blacks acquire some of the newly available homes, and a third wave of whites departs. This process continues, pro- vided there are enough blacks who want to live in the neighborhood, until even the most tolerant whites feel uncomfortable and leave.

Many studies provide direct or indirect in- formation about the link between racial com- position and neighborhood desirability. Re- searchers tend to focus on what people say or what they do. Studies that consider ex- pressed preferences usually rely on vignettes, either neighborhood diagrams or factorial surveys, to assess the effect of neighborhood traits on neighborhood desirability. Neigh- borhood diagrams present respondents with pictures of hypothetical neighborhoods and, based solely on differing racial composi- tions, ask for evaluations of each area (Farley et al. 1993; Zubrinsky and Bobo 1996). Factorial surveys also ask respondents to rate hypothetical neighborhoods, but they use verbal cues instead of pictures (Bobo and Zubrinsky 1996; Clark 1991, 1992; Schuman, Steeh, and Bobo 1985; St. John and Bates 1990). Usually racial composition is the only neighborhood trait considered in factorial surveys.

Evidence from vignette studies overwhelmingly supports the hypothesis that ra- cial composition, usually measured as per- cent black, affects neighborhood desirability. In the Detroit area, a majority of whites (87 percent) say they would move into a neigh- borhood that is 7 percent black, but only 29 percent of whites say they would select a 53 percent black neighborhood (Farley et al. 1994). In Los Angeles, the average white prefers a neighborhood that is 75 percent white and 25 percent black, but almost no whites want to live in a neighborhood that is more than 60 percent black (Clark 1992). In Oklahoma City, whites rate neighborhoods less favorably as percent black increases, even when crime, neighborhood cleanliness, neighborhood cohesion, and distance from downtown are held constant (St. John and Bates 1990).

Information about the determinants of neighborhood desirability also emerges from studies of mobility behavior. One variant of this approach uses transition matrices to evaluate mobility as a single event (Gramlich, Laren, and Sealand 1992; Massey, Gross, and Shibuya 1994; South and Crowder 1998). Massey and his colleagues show that racial composition is an important factor in the resi- dential mobility of whites and that its effect dominates concerns about the neighborhood poverty rate. Among nonpoor whites, at least 70 percent of movers select new neighbor- hoods that are in nonpoor areas where no more than 30 percent of residents are black. The probability of moving to a neighborhood that is at least 60 percent black is almost zero for this group. Similar preferences are also evident among poor whites, although their odds of selecting a pedominantly white neighborhood are lower.

Researchers also employ regression analy- sis and distinguish factors that affect the odds of leaving one's current neighborhood from those factors that influence the choice of a destination (Frey 1979; Galster 1990; Goodman and Streitwieser 1983; Harris 1997; South and Crowder 1997b, 1998; South and Deane 1993). These models often include city characteristics, characteristics of the average surrounding suburb, and differ- ences between the city and its average sub- urb as predictors of mobility behavior.

The regression approach produces incon- sistent support for the hypothesis that people base moving decisions on neighborhood ra- cial composition. Frey (1979) reports that while whites do not give much weight to ra- cial composition when deciding whether to move, racial composition is a major factor in selecting a new neighborhood. More recent work is-split about whether racial composi- tion affects the selection of destinations (Goodman and Streitwieser 1983; Ottens- mann and Gleeson 1992), but largely sup- ports Frey's conclusion that the incidence of white mobility is unrelated to the proportion of neighbors who are black (Harris 1997; South and Crowder 1997b; South and Deane 1993; but see South and Crowder 1998).

Most research supports Bobo's contention that people base their assessments of neigh- borhood desirability on racial composition. However, it is not clear why racial composi- tion matters. Questions about the meaning of apparent racial determinants of neighborhood desirability arise in part because a comple- mentary body of work suggests that neigh- borhood desirability is related to nonracial neighborhood factors. According to this lit- erature, people choose neighborhoods in an effort to avoid low socioeconomic status (SES) neighbors and associated social prob- lems. Those with low incomes, weak attach- ments to the labor force, and low levels of education are considered undesirable neigh- bors because of their perceived lack of adher- ence to mainstream values (Auletta 1982; Jencks 1992; Katz 1989; W. J. Wilson 1987). Out-of-wedlock birth, crime, unemployment, and low levels of educational attainment are all more common in neighborhoods with high proportions of low SES residents (Jargowsky 1997; Krivo and Peterson 1996; Massey 1996; Massey and Denton 1993; Wilson 1987). In addition, concerns that poverty, job- lessness, and associated problems may be contagious lead people to protect their fami- lies from "the wrong crowd" by selecting neighborhoods with as few low SES residents as possible (Jencks and Mayer 1990).

Skogan (1990) elaborates on the relation- ship between neighborhood SES, social problems, and neighborhood decline. He identifies a process by which physical "dis- order" leads to decline. This process begins when poverty, drug use, or some other prob- lem reduces residents' abilities or desires to maintain their neighborhood. The abandoned and dilapidated buildings that result harbor rats and-garbage, povide shelter for squat- ters, increase the risk of fires, house undesir- able activities, and detract from the physical attractiveness of the neighborhood. In re- sponse, those who live close to deteriorating buildings feel less satisfied with the neigh- borhood, and some decide to move. Vacan- cies are filled by less affluent families be- cause those with the ability to select more attractive neighborhoods avoid deteriorating areas. The new residents lack the resources to maintain their homes and further decay occurs. As this process of deterioration and mobility continues, neighborhood desirabil- ity declines.

In attempting to reconcile theories that highlight nonracial determinants of neigh- borhood desirability with those that empha- size responses to racial composition, it is im- portant to consider how racial and nonracial factors interact. Unfortunately, much research fails to adopt a comprehensive ap- proach to neighborhood desirability. Many studies assess the importance of racial or nonracial neighborhood factors, rather than racial and nonracial neighborhood factors. Considering multiple factors allows one to distinguish between the pure discrimination (Bobo and Zubrinsky 1996) and racial proxy hypotheses (Clark 1992; Harris 1997; Taub et al. 1984). The former argues that racial composition matters because people are sen- sitive to their neighbors' race, while the lat- ter is essentially a story about omitted-vari- able bias. The racial proxy hypothesis main- tains that when people evaluate neighbor- hoods, racial preferences simply represent a desire to live in areas free of crime, deterio- rating buildings, ineffective public schools, and other social ills. Because of the concen- tration of many social problems in neighbor- hoods with relatively large black populations (Massey 1995; Massey and Denton 1993; Peterson and Krivo 1993), selecting a "good" environment usually means choosing a pre- dominantly white neighborhood. The racial proxy hypothesis further predicts that if models are properly specified, that is, they adequately control for neighborhood social composition and social problems, racial composition will not have a significant effect on neighborhood desirability.

Most of the cited studies provide little in- sight into whether aversion to black neigh- bors is a proxy for nonracial aspects of neighborhoods. Fortunately, some work does go beyond simply estimating the total effect of racial composition. St. John and Bates (1990) observe a significant effect of percent black on neighborhood desirability in a model that also includes measures of social climate, social problems, and location. Fur- ther support for the pure discrimination hy- pothesis appears in work by Bobo and Zubrinsky (1996) and Schuman et al. (1985)

on expressed preferences, by Frey (1979) on destination selection, and by Massey, Gross, and Shibuya (1994) and Galster (1990) on mobility behavior. However, not all work that considers racial and nonracial factors re- futes the racial proxy hypothesis. Frey (1979), Harris (1997), and Taub et al. (1984) provide clear evidence of racial proxies. Each estimates a model that regresses mobil- ity incidence or mobility intentions on per- cent black. In each case there is a large total effect of race. Next, models are reestimated with controls for a host of nonracial neigh- borhood factors. Consistent with the racial proxy hypothesis, all three studies find that the direct effect of percent black is not sig- nificant, but coefficients for nonracial factors are significant (also see Marshall 1979; South and Crowder 1997b).2

While these studies represent a significant improvement over work that fails to consider nonracial factors and thus advance our un- derstanding of the relationship between neighborhood traits and neighborhood desir- ability, they clearly do not reach a consen- sus. Discrepancies are partly a result of variations in the outcomes examined (i.e., behavior versus intentions), differences in samples (i.e., national versus one metropoli- tan area), and disagreement about the unit of analysis (i.e., census tract versus city versus county), but there are also several method- ological issues that raise questions about the results of most sociological studies of neigh- borhood desirability.

First, studies that rely on expressed prefer- ences probably measure preferences with er- ror. Stated preferences about neighborhood racial and social composition may deviate from respondents' true preferences because of social pressure (Clark 1992; Kuran 1995; Schuman et al. 1985). This leads to underes- timated effects of race and inflated nonracial effects, as people attempt to conform to norms of colorblindness. In addition to de- ceit, measurement error also affects esti- mates derived from preference studies be- cause mobility intentions are an imperfect indicator of mobility behavior. Many people

A third group of studies finds that both racial and nonracial factors are important determinants of neighborhood desirability (Morenoff and Sampson 1997; South and Crowder 1997a).

indicate dissatisfaction with their neighbor- hood and plan to move, but fail to do so (Duncan and Newman 1976; Landale and Guest 1985; Speare 1974). In part this is also a problem of omitted variables, as failure to consider life-cycle issues, employment, housing supply, and neighborhood availabil- ity leads to biased estimates of the role of ra- cial and nonracial preferences in neighbor- hood desirability (Harris 1997; Rossi 1955).

Second, behavior studies that rely on tran- sition matrices are limited by two constraints. First, transition matrices cannot be constructed for continuous variables. Infor- mation is lost when continuous data are con- verted to discrete measures, and decisions about how to partition distributions can bias results. Second. transition matrices cannot easily accommodate multiple predictors. Nested matrices can be used, but this solu- tion quickly becomes unmanageable as even a modest number of independent variables quickly produces uninterpretable results. Al- ternatively, several variables could be com- bined into a single summary measure, but this solution also suffers from the limitations of discrete variables that generally plague transition matrices.

Finally, traditional regression studies of destination selection are limited by their choice of outcome variables. One option is to examine predictors of racial composition in the tracts people select (South and Crowder 1998). While this approach is bi- ased because it ignores nonracial neighbor- hood traits, the remedy for this oversight is not apparent. Including nonracial factors as independent variables does not test the racial proxy hypothesis, and creating summary measures that account for racial and nonra- cial factors suffers from the problems that plague transition matrices. A second ap- proach assesses the odds of selecting city versus suburban destinations (Frey 1979; Goodman and Streitwieser 1983; South and Crowder 1997b). This approach is an im- provement in that it allows for a broad array of destination traits to be examined, but it fails to account for the substantial heteroge- neit~within cities and suburbs and

1991; *lba 1993; Logan et al. 1996; Massey and Denton 1993). Instead, this approach adopts a false dichotomy be- tween undesirable cities and idyllic suburbs that produces results that are often difficult to interpret. Is a family that selects a subur- ban neighborhood deciding to live near blacks or whites? What is the socioeconomic status of their new neighborhood? If these critical questions cannot be answered, then we cannot draw clear conclusions from this line of research.
A Different Approach

Regression analysis, transition matrices, and vignettes each represent ineffective methods of identifying determinants of neighborhood desirability. A different approach is clearly needed. The ideal data for this research should contain information on the character- istics of every available neighborhood, unbi- ased estimates of households' satisfaction with each alternative, and identification of the neighborhoods that households select. With such data, it would be simple to recon- struct implicit preference equations by regressing desire for a given neighborhood, or the odds that a neighborhood is selected, on neighborhood characteristics. In lieu of such data, I employ hedonic price analysis.

Hedonic price analysis uses information about the prices people pay for housing to estimate preferences for dwelling and neigh- borhood characteristics. The unit of analysis is the housing unit. The hedonic model is represented by the following equation:

where P is annual housing expenditures, D is dwelling characteristics, W is a dummy vari- able identifying white households, R is neigh- borhood racial composition, and NR is non- racial neighborhood traits. The dependent variable is logarithmic and the independent variables are linear.3 As a result, coefficients identify the percent change in annual hous- ing prices associated with a one-unit change in the corresponding independent variable.

Numerous studies have concluded that this semi-log spgcification is most appropriate for he- donic price models (Chambers 1992; Kiel and Zabel 1996; Yinger 1979). Limited testing of al- ternative specifications did not identify any that were clearly superior.

I estimate two race effects. The first is a household race effect (a2),which captures the price differential between blacks' hous- ing units and whites' housing units. There are at least three reasons to suspect household race effects. First, the quality of housing units may vary by race of occupant in ways that are not accounted for by the available data. Second, neighborhood heterogeneity and persistent housing market discrimination suggest that blacks live in the least desirable sections of any given neighborhood. Third, discrimination by landlords, real estate agents, and lenders may lead blacks to expe- rience price discrimination when they rent or purchase housing (Yinger 1995). In empiri- cal work the household race effect tends to be positive, which implies that the unob- served amenities whites enjoy dominate price discrimination (Chambers 1992; Kiel and Zabel 1996).4

In addition to the household race effect, the hedonic model also estimates a neighbor- hood race effect (a3).This effect, which de- scribes the relationship between prices and neighborhood racial composition, is an indi- cator of neighborhood racial preferences as they are currently manifested in the housing market. If a3is negative, I conclude that people pay a premium to live in neighbor- hoods with relatively few black neighbors, and thus neighborhood desirability is in- versely related to percent black.

As a method for assessing the importance of various neighborhood determinants of neighborhood desirability, hedonic price analysis has several shortcomings. First, it reports preferences in the housing market, rather than among particular subgroups of the population. Property values result from bids by all potential buyers. As such, deter- minants of neighborhood desirability cannot be estimated separately for each racial group. Second, the hedonic model reports trait prices at the intersection of the supply and demand curve^.^ These forces cannot be dis-

While I do not focus on household race ef- fects and make no effort to estimate the contribu- tions of its various components, I nevertheless in- clude race of householder in all models because, as Chambers (1992) notes, failing to do so sub- stantially biases estimates of neighborhood race effects.

This characteristic of the model could lead to

entangled to produce separate estimates of the effect of neighborhood characteristics on mobility incidence and destination selection. Instead, the net effect of traits is obtained. Third, trait prices are determined by the pref- erences of the marginal consumer, not the average consumer. Linneman (1981) recog- nizes this aspect of the housing market and concludes that even if "it is known that on average the trait 2, is undesirable, one can- not a priori conclude that 2, will possess a negative coefficient in the clearing function"

(p. 134).

Despite these shortcomings, hedonic price analysis remains a useful method for describ- ing the effect of neighborhood characteristics on neighborhood desirability. First, hedonic price analysis evaluates the importance of neighborhood characteristics independent of a respondent's previous location. Conse- quently, the method does not require the as- sumptions about cities and suburbs that plague the traditional regression approach. Second, hedonic price analysis accommo- dates discrete and continuous independent variables. No bias-inducing decisions about how to convert continuous measures are nec- essary. Third, unlike transition matrices and neighborhood diagrams, hedonic price analy- sis easily accommodates multiple predictors. Finally, hedonic price analysis uses the prices people pay for their homes to assess the im- portance of neighborhood factors, rather than relying on evaluations of hypothetical neigh- borhoods, so many of the aforementioned concerns about studies of expressed prefer-

significant bias in estimates of neighborhood race effects if institutional factors restrict blacks to a subset of neighborhoods, thereby inflating prices in black neighborhoods and producing a positive trait price for neighborhood percent black. While exclusion has certainly been prominent in the housing market historically (Massey and Denton 1993: Taeuber and Taeuber 1965). it is unlikelv


that housing discrimination causes significait bias in my estimates. Recent evidence suggests that the housing market has been opening up over the past several decades (Cutler, Glaeser, and Vigdor 1997), and that housing discrimination is less pronounced in the rental market than among owner-occupied units (Yinger 1995). Addition- ally, Chambers (1992) shows that "there is no clear evidence that exclusion has elevated the overall price of ghetto or border housing above prices in the white submarket" (p. 225).

ence do not apply. Neither politically correct

responses nor the gap between intentions and

behavior affects results. Thus, while the he-

donic model probably underestimates the

magnitude of aversion to any given trait ex-

tant among the general population, it provides

important information about trait prices un-

der current market conditions.

Economists have conducted hedonic price analyses of housing for more than 25 years, but the method has rarely been employed by sociologists (F. Wilson 1979). Economic studies tend to focus on the relationship be- tween neighborhood desirability and such factors as commuting time, environmental pollution, precipitation, temperature, tax rates, crime, and street noise (Blomquist, Berger, and Hoehn 1988; Diamond and Tolley 1982; Gyourko and Tracy 1991; Linneman 1980). Racial composition is sel- dom included in these models. The failure to connect econometrics with sociological theorv means that we have limited knowl- edge-about the relationship between racial composition and property values.

Of course not all economists ignore racial composition. Yinger (1979) reviews work on the relationship between race and housing prices and concludes that there is evidence of aversion to black neighbors and discounts to white homebuyers. More recently, Cham- bers (1992) and Kiel and Zabel (1996) have used hedonic price analysis to assess racial effects on property values. Chambers exam- ines 1975 and 1979 data for Chicago and finds inconsistent evidence of race effects. His analysis does not support the claim that blacks pay higher prices than do whites, but he identifies some conditions under which neighborhood racial composition is signifi- cant. Kiel and Zabel use 1978-1991 data from the American Housing Survey to exam- ine racial effects on the value of owner-oc- cupied dwellings in Philadelphia, Chicago, and Denver. Thev confirm Chambers's conclusion that blacks are not paying more than whites for comparable housing, and they re- port broader evidence of racial aversion than Chambers observed.

I improve on this previous research in three important ways. First, I use national data, whereas Yinger (1978, 1979), Cham- bers (1992), and Kiel and Zabel (1996) each examine data for three or fewer metropolitan

areas. This is a significant difference because

the relationship between racial composition

and property values varies by metropolitan

area (Kiel and Zabel 1996). A national

sample allows me to examine geographical

differences and make broader generalizations

than is possible with limited data.

Second, I estimate coefficients for tenure

and for regional submarkets, as well as for

the national housing market. This contrasts

with Yinger, who examines data for home-

owners in one city; Chambers, who analyzes

data for Chicago; and Kiel and Zabel, who

ignore rental units.

Third, I define neighborhoods as census tracts. While there is no consensus about how neighborhoods should be defined (White 1987), tracts are commonly used by quantitative social scientists (Brooks-Gunn et al. 1993; Gramlich et al. 1992; lMassey et al. 1994). Census tracts are defined by the

U.S. Bureau of the Census with the intention of capturing neighborhoods (U.S. Bureau of the Census 1992). In defining neighborhoods as tracts, my work differs from Chambers, who uses residence zones as the geographic unit of analysis. There are 24 residence zones in the Chicago metropolitan area, which means that the average zone includes more than 250,000 people (U.S. Bureau of the Census 1981). By contrast, the average cen- sus tract includes about 4,000 people (White 1987). Thus, tracts are less heterogeneous than zones and should facilitate more precise estimates of neighborhood effects.

Data are from the Panel Study of Income Dynamics (PSID), a longitudinal survey con- ducted annually by the Survey Research Center of the University of Michigan. Initi- ated in 1968, the PSID now includes data on 37,500 individuals who resided in one of 4,800 initial sample households, were the offspring of those individuals, or were their co-residents. Because of an initial over-sampling of low-income families, the PSID sample is not representative of the U.S. population. To correct for biases, I use PSID sample weights in all analyses.6

Standard errors are computed with a Huberl White estimator (Greene 1993; StataCorp 1995).

The PSID focuses on economic and demo- graphic behavior. Respondents are queried about sources of income, changes in family structure, the acquisition of job skills, resi- dential mobility, and myriad related issues. I examine data from a special geocoded ver- sion of the PSID that was prepared in re- sponse to growing interest in contextual ef- fects. The file contains aggregate data from the 1970 and 1980 U.S. censuses, as well as codes representing each respondent's address at 12 geographic level^.^

My analysis examines data from the 1980 PSID for dwelling units in metropolitan areas. I employ these restrictions for three rea- sons. First, not all variables are available each year. For example, the PSID does not contain a description of the respondent's dwelling (i.e., one-family house, two-family house, or apartment) in 1973, 1974, or 1982. Second, at the time this research was conducted geocodes were not available for ad- dresses in 1969, 1975, 1977, 1978, or any year after 1985. Third, most places outside metropolitan areas are not tracted (U.S. Bureau of the Census 1992), so a focus on met- ropolitan areas facilitates an analysis of tract traits.

Table 1 describes selected variables from the PSID and the 1980 census. The depen- dent variable in all analyses is annual hous- ing expenditures, which is based on each respondent's report of annual rent or house value. To obtain annual expenditures from house values, I follow previous studies (Blomquist et al. 1988; Gyourko and Tracy 1991) and apply Peiser and Smith's (1985) capitalization rates to the value of owner-oc- cupied dwellings. Once rental and owner-oc- cupied prices are in the same metric, I add annual property taxes to the annual value of owned property, and convert annual housing expenditures to 1992 dollars (U.S. Bureau of the Census 1994). One potential problem with this measure is that respondents may not have a good sense of how much their homes are worth. Recent research finds that people tend to overestimate the value of their homes. However, because this error does not vary systematically across the neighborhood traits examined, imperfect estimates of hous-

'See Hill (1992) for further information about the PSID.

ing values will not bias coefficients for neighborhood traits (Goodman and Ittner 1992; Kiel and Zabel 1996).8

The equations predicting annual housing expenditures focus on the role of neighbor- hood characteristics, but also include fea- tures of the household, dwelling unit, and geographic setting that may affect property values. Race of household head is treated as a control variable, and its coefficient mea- sures the household race effect. Dwelling controls include the number of rooms in the home, whether it is a single-family dwelling, and whether it is owner-oc~upied.~

Controls for geographic setting are region and metro- politan area population. Neighborhood mea- sures are drawn from the 1980 census. One neighborhood racial factor is examined-

The results do not differ substantively when the sample is limited to households that moved during the previous year, a group that presumably has a good sense of the market value of their homes.

One potential shortcoming of the PSID is that it provides little information about characteristics of the dwelling unit. This is problematic here only if controlling for additional housing traits would significantly alter estimates of the relationship between neighborhood traits and housing costs. An alternative to the PSID is the American Hous- ing Survey (AHS). While the AHS does contain detailed dwelling information, it provides no data for areas smaller than counties. Thus, neighbor- hoods would have to be defined as counties, a choice that would likely introduce substantial noise into neighborhood-level measures and bias their estimates toward zero.

In an effort to determine the magnitude and di- rection of the bias created by controlling for just three dwelling characteristics, I examined data from the 1996 AHS. The AHS identifies residen- tial dwelling units that are in the same "pseudo- tract." Pseudo-tracts are similar to census tracts, and on average contain 98 sample housing units

(U.S. Bureau of the Census 1998). With pseudo- tracts as proxies for neighborhoods, I used the race of sample household heads to estimate neighborhood racial composition. Next, I esti- mated hedonic models, first with the PSID dwell- ing unit characteristics and then with an addi- tional group of dwelling unit characteristics avail- able in the AHS. This exercise suggests that con- trolling for the three PSID dwelling unit charac- teristics leads to estimates of neighborhood race effects that are biased by k.04, a difference that is not statistically significant at conventional lev- els (tables available on request).
1980 U.S. Census
Variable     Mean Deviation
PSID Measures     
Annual housing expenditure     

Number of rooms

Single-family house







Cerzsus Tract Measures, 1980

Metropolitan area population

Percent black

Percent affluent

Percent poor

Percent with no college

Percent unemployed

percent black. Four nonracial factors, all components of neighborhood socioeconomic status, are also considered: percent affluent, percent poor, percent with no college educa- tion, and percent unemployed.

Because of a concern that the effect of ra- cial composition on desirability is a function of percent black (Galster 1990; Schelling 1971; Yinger 1979), I use dummy variables to measure neighborhood racial composi-


Renters: "About how much rent do you pay a month?" Monthly rent was multiplied by 12. Own- ers: "Could you tell me what the present value of your home is-I mean about what would it bring if you sold it today?" These values were multiplied by a capitalization rate of ,0785 and added to estimated annual property taxes. Annual housing expenditures are expressed in 1992 dollars.

"How many rooms do you have (for your family), not counting bathrooms?"

"Do you live in a one-family house, a two-family house, an apartment, or what?" Respondents in one- family homes are coded 1, otherwise 0.

Coded 1 if the respondent is a homeowner, other- wise 0.

Coded 1 for respondents in the Northeast, otherwise 0.

Coded 1 for respondents in the Midwest, otherwise 0.

Coded 1 for respondents in the West, otherwise 0.

Coded 1 for respondents in the South, otherwise 0.

Coded 1 for non-Hispanic white household heads, otherwise 0.

Metropolitan area population (in 100,000s) in 1980.

Percent black in 1980 census tracts.

Families reporting annual incomes greater than $30,000 ($5 1,142 in 1992 dollars).

Percent of residents in households with annual in- comes below the poverty line.

Percent of residents at least 25 years old who have no more than 12 years of schooling.

Percent of residents at least 16 years old who are in the civilian labor force and unemployed.

tion.1° Each dummy variable represents a segment of the distribution of percent black. After considering several coding schemes, I divided the distribution into three segments: less than 10 percent black, 10 percent to 60

1°1also considered splines and a quadratic term as methods of capturing nonlinear racial effects. Dummy variables fit the data better than did ei- ther alternative.

Table 2. Percentage Distribution of Census-Tract Percent Black and Mean Annual Housing Expen- diture by Tenure and Region, 1980

Census-Tract Percent Black

Less than 10 to 60 At Least Variable 10 Percent Percent 60 Percent Total


Unweighted N

Percentage (weighted)

Mean annual housing expenditure

By Tenure

Renters: Unweighted N Percentage (weighted) Mean annual housing expenditure

Homeowners: Unweighted N Percentage (weighted) Mean annual housing expenditure

By Region


Unweighted N

Percentage (weighted)

Mean annual housing expenditure

Midwest: Unweighted N Percentage (weighted) Mean annual housing expenditure

West: Unweighted N Percentage (weighted) Mean annual housing expenditure

South: Unweighted N Percentage (weighted) Mean annual housing expenditure

percent black, and at least 60 percent black. ing a reduced model that includes all controls These ranges are consistent with those used and independent variables, with the excep- by other researchers (Chambers 1992; Kiel tion of the nonracial neighborhood traits. and Zabel 1996). Also, tests of other ranges Second, I explore the racial proxy argument of the percent black distribution produced by adding nonracial factors to the reduced none that fit the data significantly better. model and estimating the direct effect of ra- Table 2 presents the distribution of respon- cial composition on property values." Third, dents across categories of percent black in the census tract.

" This is not the true direct effect because crime, school quality, and other potentially im- RESULTS portant neighborhood traits are not included in the model. AS a result, if any neighborhood race

The proceeds in three parts. First, I

effect remains after controlling for neighborhood assess the total effect of racial ~0mp0~iti0n SES. I cannot definitivelv conclude that this is

on annual housing expenditures by estimat- evidence of pure discrimination. Some or all of

Table 3. Unstandardized Coefficients from the Regression of Annualized Housing Expenditures on Selected Independent Variables: All Households, 1980
Model 1     Model 2
Independent Variable     Coefficient     S.E.     Coefficient     S.E.
Percent black:         

Between 10 and 60 percent -.163"* (.040) -.023 (.035) At least 60 percent -.460 ' (.057) -.093 (.059)

U7hite household Percent affluent Percent poor Percent with no college Percent unemployed

Number of rooms Single-family dwelling Homeowner

Metropolitan area population ,005" (.001) .003** (.001)

Northeast Midwest West

Constant 7.286"" (.061) 7.832" (.100) Adjusted R2 .49 .58

*p< .05 '-p < .O1 (two-tailed tests)

I evaluate both models separately for rental, to a $1,187 reduction in annual costs. An owner-occupied, and regional housing even larger reduction in housing value is as- submarkets to ascertain whether the impor- sociated with moving from a neighborhood tance of racial and nonracial neighborhood in which less than 10 percent of residents traits varies by tenure or region. are black to one where at least 60 percent of

Table 3 reports coefficients from the re- the neighborhood is black. In this scenario, duced model, Model 1, for all dwelling dwellings lose 46 percent of their annual units. Racial composition is the only neigh- value, which is equivalent to about a $3,351 borhood trait included in the model.12 He- reduction in annual housing costs for the av- donic coefficients indicate that housing erage unit. These race effects are highly sig- units lose about 16 percent of their value nificant, both statistically and substantively, when neighborhood racial composition in- and are consistent with the observation that creases from less than 10 percent black to "property values drop when black families between 10 percent and 60 percent black. move in" (Farley et al. 1994:775). For the average dwelling, this is equivalent Model 1 in Table 3 provides a strong affir-

mative response to the question, "Are hous- the surviving racial aversion may yet be a proxy ing prices affected by the proportion of local for nonracial factors.

neighbors who are black?" However, it of-

l2 In this and all subsequent equations, tracts

fers no evidence about why racial composi-

less than 10 percent black are the omitted cat-

tion matters. To address the issue of motiva-

egory. As a result, the two dummy variables for

tion, I consider whether sensitivity to black

percent black indicate how much annual housing

neighbors can be explained by a desire to

costs would change if the unit moved from one of the least integrated neighborhoods to an area with live in high SES neighborhoods. Model 2 in a higher concentration of black residents. Table 3 assesses the effect of racial and non-

racial neighborhood traits on annual housing expenditures by including all predictors in one equation.13

Coefficients for the nonracial neighbor- hood traits indicate that these factors play an important part in determining property values. Each coefficient is in the hypoth- esized direction, and all are statistically sig- nificant. For each percentage-point increase in the percentage of affluent neighbors, an- nual housing costs increase by .63 percent. For the average home, the effect of a mar- ginal increase in a neighborhood's percent affluent is a $46 annual increase in housing costs. Similarly, increases in the three low- status neighborhood measures lead to reduc- tions in property values. The magnitude of these effects for percent poor, percent with no college, and percent unemployed is in the range of -.63 percent to -.88 percent, implying that the average home is worth be- tween $46 and $64 less per year when neighborhood socioeconomic status experi- ences a marginal decline.

Model 2 in Table 3 also provides insight into why people avoid black neighbors. There is strong evidence that lower housing costs in more integrated neighborhoods are primarily a response to the neighborhood's socioeconomic status rather than the race of its residents. After controlling for neighbor- hood SES, there is no longer a significant relationship between racial composition and annual housing costs.14 Compared to the re- duced model, estimates based on this more complete model show at least an 86 percent reduction in the monetary consequences of integration. Evidence of a preference for high SES neighbors, combined with the lack of a significant net effect of percent black, provides strong support for the racial proxy hypothesis. Clearly, housing is more valu- able in less integrated neighborhoods

l3 An examination of correlation coefficients shows that while some relationships are strong, as high as k.74, my findings do not result from multicollinearity. Standard errors for the percent black variables change little between the reduced and full models, and substantial reductions in neighborhood race effects can be obtained by controlling for only one nonracial trait at a time.

l4 An F-test supports the hypothesis that the ra- cial composition coefficients jointly equal zero (F2,4132= 1.25,p = .29).

largely because people prefer well-educated, affluent neighbors, and each of these traits is more prevalent among whites than among blacks.

The results so far assume that the processes determining housing prices do not vary by region or tenure. There are at least four rea- sons why this assumption may be incorrect. First, there are consistent significant effects of tenure and region on annual housing costs. In the full model, people pay 41 percent more to own a home, and at least 8 percent more if their residence is located in the West or Northeast. Second, mean annual housing expenditures vary across tenure and regional subgroups (Table 2). Third, several recent studies find significant variation by region in whites' racial preferences (Clark 1991), the stability of integration (Lee and Wood 1991), and segregation levels (Farley and Frey 1994). Fourth, renters and owners do not have equal stakes in their neighborhoods. To an owner, a home is both a residence and an investment; to a renter, a home is a place to live. This difference means that, compared with renters, owners give greater weight to future property values when determining a dwelling's current value.

To allow trait prices to vary by tenure and region, I estimate the reduced and full mod- els separately for renters, owners, and each of the four regions. Table 4 reports findings for regional submarkets. Model 1 excludes neighborhood SES. Consistent with research that reports lower levels of residential segre- gation in the West (Farley and Frey 1994), coefficients for racial composition in the re- duced model are least negative in this region: Western housing loses no more than 33 percent of its value when located in neighbor- hoods that are more than 10 percent black. By contrast, reductions in annual costs are as much as 40 percent in the South, 52 percent in the Midwest, and 70 percent in the North- east for dwellings located in neighborhoods that are more than 10 percent black. l5

l5 Only three race effects differ significantly between regions. They are between the Northeast and West, Midwest and West, and West and South for 10 percent to 60 percent black tracts.

Table 4. Unstandardized Coefficients from the Regression of Annualized Housing Expenditures on Selected Independent Variables, by Region, 1980
Model 1         Model 2
Independent Variable     Coefficient     S.E.     Coefficient     S.E.
Percent black:                     
Between 10 and 60 percent     -.210*     (.092)     ,010         (.084)
At least 60 percent     -.695*'     (.189)     -.214         (.176)
White household                     
Percent affluent                     
Percent poor                     
Percent with no college                     
Percent unemployed                     
Adjusted R2     .46             .59     
Percent black:                     
Between 10 and 60 percent     -.193*     (.077)     -.I14         (.068)
At least 60 percent     -.517**     (.119)     -. 132         (.128)
White household                     
Percent affluent                     
Percent poor                     
Percent with no college                     
Percent unemployed                     
Adjusted R2     .5 1             .61     
Percent black:                     
Between 10 and 60 percent     ,142     (.081)     ,205"         (.067)
At least 60 percent     -.331"     (.094)     -.I57         (.102)
White household                     
Percent affluent                     
Percent poor                     
Percent with no college                     
Percent unemployed                     
Adjusted R2     .57             .65     
Percent black:                     
Between 10 and 60 percent     -.241*'     (.062)     -. 105         (.055)
At least 60 percent     -.399*'     (.093)     -.020         (.098)
White household                     
Percent affluent                     
Percent poor                     
Percent with no college                     
Percent unemployed                     
Adjusted R2     .44             .5 1     

Note: Models also include all controls, except controls for region. *p< .05 "p < .O1 (two-tailed tests)

Model 2 in Table 4 adds neighborhood SES variables to the previous equation to address the question of why aversion to blacks appears in regional models. As for the national sample, controlling for nonra- cial factors results in dramatic reductions in the neighborhood race effect: The full model shows no evidence of a significant aversion to black neighbors in any region.16 Instead, neighbors' income, employment status, and educational attainment are the significant neighborhood-level determinants of property values in all regions. These re- sults provide further support for the racial proxy hypothesis.17

Table 5 presents results separately for rent- ers and homeowners. Model 1 reports a negative relationship between racial compo- sition and prices in both submarkets. Three of the four neighborhood race coefficients are statistically significant, and the cost of integration is greater in neighborhoods with a heavier concentration of black residents. However, responses to racial composition are not comparable for renters and owners; they differ in the magnitude of the response to ra- cial composition. The annual expense of rental property is only significantly lower when percent black exceeds 60 percent, in which case costs decline by 31 percent com- pared to similar units in tracts that are less than 10 percent black. By contrast, owner- occupied units lose between 22 percent and 59 percent of their value when the percent- age of black neighbors exceeds 10 percent. This difference in sensitivity to racial com- position, which is statistically significant in neighborhoods that are at least 60 percent black, supports the hypothesis that home-

l6 After controlling for neighborhood SES, sig- nificant regional differences remain between the Midwest and West, and West and South for 10 percent to 60 percent black tracts.

l7 The significant positive coefficient for West- ern residences located in 10 percent to 60 percent black census tracts indicates either that there is a substantial taste for integration in this region, or that these neighborhoods contain unobserved amenities that cause prices to be higher than in less integrated areas. However, this result is not a "California effect." Although 56 percent of West- ern sample households are in California, control- ling for California residence has little effect on the coefficients for percent black.

owners are more deeply invested in their homes than are renters.

Model 2 in Table 5 reports coefficients from the full model. Once again, tenure sta- tus affects neighborhood trait prices. Rental prices are sensitive to the local poverty rate and the percentage of residents who have not attended college. Neither percent affluent nor percent unemployed has a significant impact on the annual cost of rental housing. By con- trast, purchase prices are significantly related to all neighborhood SES variables except percent poor. Model 2 also shows differences in why renters and owners pay more to live near relatively few blacks. For renters, the full model offers compelling evidence of ra- cial proxies. After controlling for neighbor- hood SES, the effect of percent black on rental prices is small and not significantly different from zero. Support for the racial proxy hypothesis is inconsistent for homeowners. While each coefficient for per- cent black is about 65 percent less negative in the full model than in Model 1, racial composition continues to be a significant predictor of prices in neighborhoods that are at least 60 percent black. Homes located in these predominantly black areas, as opposed to neighborhoods that are less than 10 per- cent black, lose 21 percent of their annual value after controlling for neighborhood SES. This finding is consistent with the pure discrimination hypothesis and suggests that, possibly because of the sizable investment associated with owning a home, buyers in predominantly black neighborhoods are con- cerned about the race of their neighbors and not just their social class.

In failing to reject the pure discrimination hypothesis, results for owner-occupied units in predominantly black neighborhoods differ from those presented for the national hous- ing market, rental and regional submarkets, and the remainder of the owner-occupied submarket. What proportion of these dwell- ing units are evaluated based on the race, per se, of local residents? Among the sample of 1980 housing units I have examined, only 4 percent are owner-occupied and located in tracts that are at least 60 percent black. Thus, although pure discrimination cannot be re- jected as an accurate description of why ra- cial composition affects prices in all neigh- borhoods, my analysis shows that for 96 per-

Table 5. Unstandardized Coefficients from the Regression of Annualized Housing Expenditures on Selected Independent Variables, by Tenure, 1980

Independent Variable


Percent black: Between I0 and 60 percent At least 60 percent

White household Percent affluent Percent poor Percent with no college Percent unemployed

Adjusted R2


Percent black: Between 10 and 60 percent At least 60 percent

White household Percent affluent Percent poor Percent with no college Percent unemployed

Adjusted R2 Model 1 Model 2 Coefficient S.E. Coefficient S.E.

Note: Models also include all controls, except the control for tenure.

"p< .05 v'p < .O1 (two-tailed tests)

cent of housing units, price declines as neighborhood percent black increases be- cause of neighbors' socioeconomic status, not their race.

I address two questions. Are housing prices lower in neighborhoods with high concentra- tions of black residents? If so, is this rela- tionship evidence of pure discrimination or can it be explained by considering nonracial neighborhood traits? These questions derive their importance from the link between mo- bility patterns and racial residential segrega- tion, and the subsequent relationship be- tween high levels of residential segregation and myriad social problems (Cutler and Glaeser 1997; Massey and Denton 1993; Massey and Eggers 1990; Massey, Gross, and Eggers 1991; Massey et al. 1994). More- over, determining whether the pure discrimi- nation or racial proxy hypotheses best de- scribes the relationship between racial com- position and neighborhood desirability is im- perative because the two explanations have disparate implications for integration policy.

Although previous work estimates aver- sion to black neighbors, few studies have ex- plored the motivations for this bias. Instead, most work has focused on racial explanations and has failed to consider the racial proxy hypothesis. Much of previous work uses vi- gnette, regression analysis, or transition ma- trix approaches, all of which have a limited ability to produce unbiased estimates of the effects of racial and nonracial neighborhood traits on neighborhood desirability. Instead, I have employed hedonic price analysis, a powerful tool for describing the relationship between prices and neighborhood composi- tion under current housing market condi- tions. This method also provides indirect in- formation about whites, as their numerical superiority in most housing markets means that trait prices largely reflect whites' prefer- ences.

My assessment of housing prices yields three main conclusions. First, with respect to the national market, property values do re- spond to racial composition. Housing loses at least 16 percent of its value when located in neighborhoods that are more than 10 per- cent black. This finding corroborates previ- ous sociological research showing that whites, but also members of other racial and ethnic groups, prefer not to have black neigh- bors (Bobo and Zubrinsky 1996; Clark 1991, 1992; Farley et al. 1994; St. John and Bates 1990; Zubrinsky and Bobo 1996).

Second, for the nation as a whole, my work provides insight into how this observed aver- sion to black neighbors should be inter- preted. I show that housing in neighborhoods with a high percentage of black residents is less valuable not because of an aversion to blacks per se, but rather because people pre- fer affluent, well-educated neighbors, and these traits are more common among whites than blacks. This finding strongly supports the racial proxy hypothesis. In addition, it begs reconsideration of previous work on the relationship between racial composition and neighborhood desirability. My work confirms Chambers's (1992) finding that esti- mates of the effect of neighborhood racial composition that do not control for nonracial factors are biased upward, and extends this conclusion beyond the Chicago context Thus, much previous work probably overstates the effect of racial composition on neighborhood desirability.

Third, there are important submarket dif- ferences in the magnitude and motivations of price sensitivity to racial composition. Re- sponsiveness to black neighbors is most ex- treme among homeowners and in regions out- side the West. Homeowners require a 59 per- cent price reduction as compensation for liv- ing in a neighborhood that is at least 60 per- cent black. Aversion to blacks in the owner- occupied submarket is less clearly motivated by a preference for high SES neighbors. Even after controlling for nonracial neighborhood traits, owner-occupied units are still worth 21 percent less in neighborhoods that are at least

60 percent black than when they are located in predominantly white areas. Therefore, in contrast to the national sample, regional subsamples, and the rental subsample, results for a small segment of the owner-occupied submarket are-consistent with the pure-dis- crimination hypothesis. A synthesis of these conclusions is that "property values drop when black families move in" (Farley et al. 1994:775), but whether it is blacks' race or their class that determines property values depends on whether dwellings are rented or owner-occupied and the percentage of current residents that are black.

My work offers a cautiously optimistic message for integration policy. The conclu- sion that people generally avoid black neigh- bors for reasons that are related to social class bodes well for stable integration. When black residents and their nerghbors have similar socioeconomic statuses, increasing levels of integration should have little effect on property Jalues, and white flight should not ensue. However, my work also suggests that stable integration may not be an attain- able goal in all contexts. If home prices in predominantly black neighborhoods are in- deed responsive to racial composition, then integration may lead to racial succession in these areas.

My work identifies several clear directions for future research. To definitively conclude that either the racial proxy or pure discrimi- nation hypotheses accurately describe the re- lationship between racial composition and neighborhood desirability, models must evaluate a broader range of neighborhood traits. Crime and school quality are two fac- tors that might account for the residual ef- fect of racial composition in the owner-oc- cupied submarket. Also, prices may be re- sponsive to changes in neighborhood traits, not just their levels. Unfortunately, data limi- tations prevented consideration of these fac- tors. but an examination of additional static and'dynamic neighborhood traits is planned. Future research also should address how members of various racial groups differ in their conceptions of the relationship between racial composition and neighborhood desir- ability. Even though the effect of neighbor- hood racial composition on housing prices is explained by social composition for most dwellings, aversion to blacks for racial rea- sons may vary across racial groups. Finally, future work could extend the current analy- sis by broadening its conception of racial composition. As the nation's racial diversity increases, we need to understand not only the effect of black residents on neighborhood desirability, but the effects of Latino and Asian neighbors as well.

David R. Harris is Assistant Professor of Sociol- ogy at the University of Michigan. His current research projects examirie racial differences in suburbanization, the role of racial and nonracial neighborhood traits in whites' decisions to rnove, determinants offirm relocation decisions, and the educational and social development of mixed- race clzildren. He is also involved in the develop- ment of the Los Angeles Survey of Families and Communities.

Alba, Richard D. and John R. Logan. 1991. "Variations on Two Themes: Racial and Eth- nic Patterns in the Attainment of Suburban Residence." Demography 28:43 1-53.

Auletta, Ken. 1982. The Uriderclass. Nen York: Vintage Books. Blomquist, Glenn C., Mark C. Berger, and John

P. Hoehn. 1988. "New Estimates of Quality of Life in Urban Areas." American Economic Re- view 78:89-107.

Bobo, Lawrence D. 1997. "The Color Line, the Dilemma, and the Dream." Pp. 31-55 in Civil Rights and Social Wrongs, edited by J. Higham. University Park, PA: Pennsylvania State University Press.

Bobo, Lawrence, James R. Klugel, and Ryan A. Smith. 1997. "Laissez-Faire Racism: The Crys- tallization of a Kinder, Gentler, Antiblack Ide- ology." Pp. 15-42 in Racial Attit~ides in the 1990s, edited by S. A. Tuch and J. K. Martin. Westport, CT: Praeger.

Bobo, Lawrence and Camille L. Zubrinsky. 1996. "Attitudes on Residential Integration: Per- ceived Status Differences, Mere In-Group Pref- erence, or Racial Prejudice." Social Forces 74:883-909.

Brooks-Gunn, Jeanne, Greg J. Duncan, Pamela Kato Klebanov, and Naomi Sealand. 1993. "Do Neighborhoods Influence Child and Adolescent Development?" American Journal Of Sociol- ogy 99:353-95.

Chambers, Daniel N. 1992. "The Racial Housing Price Differential and Racially Transitional Neighborhoods." Journal of Urban Econornics 32:214-32.

Clark, William A. V. 1991. "Residential Prefer- ences and Neighborhood Racial Segregation: A Test of the Schelling Segregation Model." Denzography 28:l-19.

. 1992. "Residential Preferences and Resi- dential Choices in a Multiethnic Context." Deriiography 29:45 1-66.

Cutler, David M, and Edward L. Glaeser. 1997. "Are Ghettos Good or Bad?" The Quarterly Journal of Econor~zics 112:827-72.

Cutler, David M., Edward L. Glaeser, and Jacob

L. Vigdor. 1997. "The Rise and Decline of the American Ghetto." Working Paper 5881, Na- tional Bureau of Economic Research, Cam- bridge, MA.

Diamond, Douglas B., Jr. and George S. Tolley. 1982. The Ecorzo~nics of Urban Amenities. New York: Academic Press.

Duncan, Greg J. and Sandra J. Newman. 1976. "Expected and Annual Residential Mobility." Journal of the American Instit~ite of Planners

42: 174-86.

Farley, Reynolds and William H. Frey. 1994. "Changes in the Segregation of Whites from Blacks During the 1980s: Small Steps toward a More Integrated Society." American Sociologi- cal Review 59:23-45.

Farley, Reynolds, Charlotte Steeh, Tara Jackson, Maria Krysan, and Keith Reeves. 1993. "Con- tinued Racial Residential Segregation in De- troit: 'Chocolate City, Vanilla Suburbs' Revis- ited." Journal of Housing Research 4:l-38.

Farley, Reynolds, Charlotte Steeh, Maria Krysan, Tara Jackson, and Keith Reeves. 1994. "Ste- reotypes and Segregation: Neighborhoods in the Detroit Area." American Journal of Sociol- ogy 100:750-80.

Frey, William H. 1979. "Central City White Flight: Racial and Nonracial Causes." American Sociological Review 44:425-48.

Galster, George C. 1990. "White Flight from Ra- cially Integrated Neighborhoods in the 1970s: The Cleveland Experience." Urban Studies 27:385-99.

Goodman, John L., Jr. and John B. Ittner. 1992. "The Accuracy of Home Owners' Estimates of House Value." Journal of Housing Economics 2:339-57.

Goodman, John L., Jr. and Mary L. Streitwieser. 1983. "Explaining Racial Differences: A Study of City-to-Suburb Residential Mobility." Urban Affairs Quarterly 18:301-25.

Gramlich, Edward, Deborah Laren, and Naomi Sealand. 1992. "Moving Into and Out of Poor Urban Areas." Journal of Policy Analysis and Marzageme~zt11:273-87.

Greene, William H. 1993. Ecoaonzetric Analysis. Englewood Cliffs, NJ: Prentice Hall.

Gyourko, Joseph and Joseph Tracy. 1991. "The Structure of Local Public Finance and the Quality of Life." Journal of Political Economy 99:774-806.

Harris, David R. 1997. "Race, Class, and Social Problems: Determinants of White Residential Mobility." Ph.D, dissertation, Department of Sociology, Northwestern University, Evanston, IL.

Hill, Martha S. 1992. The Panel Study of Income Dyrzanzics: A User's Gllide. Newbury Park, CA: Sage.

Jargowsky, Paul A. 1997. Poverty and Place: Ghettos, Barrios, and the American City. New York: Russell Sage Foundation.

Jencks, Christopher. 1992. Rethinking Social Policy: Race, Poverty, and the Urzderclass.

Cambridge, MA: Harvard University Press.

Jencks, Christopher and Susan E. Mayer. 1990. "The Social Consequences of Growing Up in a Poor Neighborhood." Pp. 11 1-86 in Inner-City Poverty in the United States, edited by L. E. Lynn Jr, and M. G. H. McGeary. Washington, DC: National Academy Press.

Katz, Michael B. 1989. The Undeserving Poor: Fivriz the War or1 Poverty to the War on Wel- fare. New York: Pantheon Books.

Kiel, Katherine A. and Jeffrey E. Zabel. 1996. "House Price Differential in U.S. Cities: Household and Neighborhood Racial Effects." Jouriznl of Housing Economics 5: 143-65.

Krivo, Lauren J. and Ruth D. Peterson. 1996. "Extremely Disadvantaged Neighborhoods and Urban Crime." Social Forces 75:619-48.

Kuran, Timur. 1995. "The Inevitability of Future Revolutionary Surprises." American Journal of Sociology 100: 1528-5 1.

Landale, Nancy S. and Avery M. Guest. 1985. "Constraints, Satisfaction, and Residential Mo- bility: Speare's Model Reconsidered." Demograp$ 22: 199-222.

Lee, Barrett A. and Peter B. Wood. 1991. "Is Neighborhood Racial Succession Place-Spe- cific?" Deritography 28:21-40.

Linneman, Peter. 1980. "Some Empirical Results on the Nature of the Hedonic Price Function for the Urban Housing Market." Journal of Urban Ecoriornics 8:47-68. . 1981. "The Demand for Residence Site Characteristics." Journal of Urban Econontics 9:129-48.

Logan, John R. and Richard D. Alba. 1993. "Locational Returns to Human Capital: Minor- ity Access to Suburban Community Resources." Dernogmphy 30:243-68.

Logan, John R., Richard D. Alba, Tom McNulty, and Brian Fisher. 1996. "Making a Place in the Metropolis: Locational Attainment in Cities and Suburbs." Denzography 33:443-53.

Marshall, Harvey. 1979. "White Movement to the Suburbs: A Comparison of Explanations." Anzerican Sociological Review 44:975-94.

Massey, Douglas S. 1995. "Getting Away with Murder: Segregation and Violent Crime in Ur-

ban America." University of Pennsylvania Law Review 143:1203-32.

. 1996. "The Age of Extremes: Concen- trated Affluence and Poverty in the Twenty- First Century." Dentogrphy 33:395-412.

Massey, Douglas S. and Nancy A. Denton. 1993.

Anzericaiz Apartheid: Segregation and the Making of the Underclass. Cambridge, MA: Harvard University Press.

Massey, Douglas S, and Mitchell L. Eggers. 1990. "The Ecology of Inequality: Minorities and the Concentration of Poverty, 1970-1980." American Journal of Sociology 95:1153-88.

Massey, Douglas S., Andrew B. Gross, and Mitchell L. Eggers. 1991. "Segregation, the Concentration of Poverty, and the Life Chances of Individuals." Social Science Research 20: 397-420.

Massey, Douglas S., Andrew B. Gross, and Kumiko Shibuya. 1994. "Migration, Segrega- tion, and the Geographic Concentration of Pov- erty." American Sociological Review 59:425


Morenoff, Jeffrey D. and Robert J. Sampson. 1997. "Violent Crime and the Spatial Dynam- ics of Neighborhood Transition: Chicago, 1970-1990." Social Forces 76:31-64.

Ottensmann, John R. and Michael E. Gleeson. 1992. "The Movement of Whites and Blacks into Racially Mixed Neighborhoods: Chicago, 1960-1980." Social Science Quarterly 73:645


Peiser, Richard B, and Lawrence B. Smith. 1985. "Homeownership Returns, Tenure Choice, and Inflation." American Real Estate and Urban Economics Journal 13:343-60.

Peterson, Ruth D. and Lauren J. Krivo. 1993. "Racial Segregation and Black Urban Homi- cide." Social Forces 71: 1001-26.

Rossi, Peter H. 1955. Why Families Move. New York: Free Press.

Schelling, Thomas. 1971. "Dynamic Models of Segregation." Journal of Mathematical Sociol- ogy 1: 143-86.

Schuman, Howard, Charlotte Steeh, and Lawrence Bobo. 1985. Racial Attitudes in America: Trends and Interpretations. Cambridge, MA: Harvard University Press.

Skogan, Wesley G. 1990. Disorder and Decline. New York: Free Press.

South, Scott J. and Kyle D. Crowder. 1997a. "Escaping Distressed Neighborhoods: Indi- vidual, Community, and Metropolitan Influ- ences." Anterican Journal of Sociology 102:1040-84. . 1997b. "Residential Mobility between Cities and Suburbs: Race, Suburbanization, and Back-to-the-City Moves." Demography 34: 525-38.

, 1998. "Leaving the 'Hood: Residential Mobility between Black, White, and Integrated Neighborhoods." American Sociological Re- view 63:17-26.

South, Scott J. and Glenn D. Deane. 1993. "Race and Residential Mobility: Individual Determi- nants and Structural Constraints." Social Forces 72: 147-67.

Speare, Alden, Jr. 1974. "Residential Satisfaction as an Intervening Variable in Residential Mo- bility." Demography ll: 173-88.

St. John, Craig and Nancy A. Bates. 1990. "Ra- cial Composition and Neighborhood Evalua- tion." Social Science Research 19:47-61.

StataCorp. 199.5. Stata Reference Manual. Release 4.0. College Station, TX: Stata Corpora- tion.

Taub, Richard P., Garth D. Taylor, and Jan D. Dunham. 1984. Paths of Neighborhood Change. Chicago, IL: University of Chicago Press.

Taeuber, Karl E. and Alma F. Taeuber. 1965. Negroes in Cities: Residential Segregation and Neighborhood Change. Chicago, IL: Aldine Publishing.

U.S. Bureau of the Census. 1981. Census of Population and Housing, 1980; Characteristics of the Population (PC80-1-B15). Washington, DC: U.S. Government Printing Office.

. 1992. Census of Population and Housing, 1990: Summary Tape File 3. Washington, DC:

U.S. Government Printing Office.

. 1994. Statistical Abstract of the United States: 1994. Washington, DC: U.S. Govern- ment Printing Office.

. 1998. American Housing Surlley for the United States. Washington, DC: U.S. Govern- ment Printing Office.

White, Michael J. 1987. American Neighbor- hoods and Residential Differentiation. New York: Russell Sage Foundation.

Wilson, Franklin D. 1979. Residential Consump- tion, Economic Opportunity, and Race. New York: Academic Press.

Wilson, William Julius. 1987. The Trul? Disad- vantaged: The Inner City, the Underclass, and Public Policy. Chicago, IL: University of Chi- cago Press.

Yinger, John. 1978. "The Black-White Price Dif- ferential in Housing: Some Furthcr Evidence." Land Econontics 54: 187-206. . 1979. "Prejudice and Discrimination in the Urban Housing Market." Pp. 430-68 in Current Issues in Urban Economics, edited by

P. Mieszkowski and M. Straszheim. Baltimore, MD: Johns Hopkins University Press.

. 1995. Closed Doors, Opportunities Lost: The Continuing Costs of Housing Discrimina- tion. New York: Russel Sage Foundation.

Zubrinsky, Camille L. and Lawrence Bobo. 1996. "Prismatic Metropolis: Race and Residential Segregation in the City of Angels." Social Sci- ence Research 25:335-74.

  • Recommend Us