Document View

Historically, public and other federally assisted housing developments have contributed to urban racial segregation. In the 1970s, however, HUD adopted regulations that discouraged the location of assisted housing developments in areas with high percentages of minority households. This article looks at the role of race, ethnicity, and poverty in the siting of five types of assisted housing during the 1980s. HUD data are combined with census data to identify the characteristics of the tracts that received public and assisted housing. Although the value of the owner occupied units in tracts was the strongest predictor of the placement of most types of assisted housing, the results indicate that race and ethnicity still mattered.

Due to its historic placement in areas of minority concentration, assisted housing has been criticized for contributing to residential segregation (Galster, 1999; Massey & Denton, 1993; Massey & Kanaiaupuni, 1993). Galster (1999), for example, refers to "a disgraceful legacy of blatant discrimination in the operation of our public housing programs" (p. 125). During the 1970s, however, the U.S. Department of Housing and Urban Development (HUD) instituted siting guidelines designed to discourage local housing authorities from concentrating new HUD-funded housing developments in neighborhoods with high proportions of minority residents. It is not clear, however, whether race continues to be an important determinant in the siting of subsidized housing developments.

Criticisms of assisted housing with regard to its impact on segregation have largely been directed toward the public housing program (Bauman, 1987; Hirsch, 1983; Massey & Kanaiaupuni, 1993). Over the last few decades, however, new assisted housing developments have been funded through other programs such as Section 8 New Construction and the Low Income Housing Tax Credit (LIHTC) programs. Yet we know very little about the characteristics of neighborhoods in which these other types of assisted housing developments have been sited. In this article, we consider the siting of family and elderly public housing, family and elderly other HUD housing, and LIHTC housing.

Changes in the demographics of American society also warrant a reexamination of the link between the siting of assisted housing and minority concentration. The focus of most prior studies of the siting of assisted housing was on whether it was being built in African American neighborhoods. But Latinos and Asians are now a major presence in urban America. Therefore, in this article we examine the relationship between concentrations of Latino and Asian residents and the siting of assisted housing.

Our intent is to assess the role of race and ethnicity in the siting of assisted housing during the 1980s. We are interested in finding whether the historic pattern of placing assisted housing in areas of minority concentration has changed. To achieve this purpose, we analyze the siting patterns during the 1980s of five types of assisted housing. Using data from the U.S. Census and HUD's Picture of Subsidized Housing data set, we identify the social and physical characteristics of neighborhoods in which subsidized housing was sited. We also assess the relative importance of each neighborhood characteristic and explore the implications of this analysis for both social theory and urban policy.

The Siting of Assisted Housing

Two theoretical perspectives have been applied to understanding the siting of assisted housing: the political economy of race perspective and the urban ecology perspective. The political economy of race perspective on urban spatial structure points to racism and the choices made by local elites as the key factors in determining the landscape of urban America (Goldstein & Yancey, 1986). Because public and other forms of assisted housing are generally viewed as undesirable additions to a neighborhood, the weakest groups in society have borne their burden. The decentralized nature of the public housing program gave local elites the power to site public housing in neighborhoods that would mount the least resistance. Schill and Wachter (1995) note that the Housing Act of 1937 instituted a structure in which local public housing authorities, rather than the federal government, were given the authority to locate public housing units. Many local governments used this authority to place public and assisted housing in poor, predominantly African American neighborhoods.

African Americans, and more recently Latinos, have been among the politically weakest groups in urban America. It is not surprising, then, that racism and the power held by local political regimes can explain the historic placement of public housing in low-income, minority areas (Bauman 1987; Galster & Keeney, 1993; Goering, 1994; Goldstein & Yancey, 1986; Massey & Kanaiaupuni, 1993; Myerson & Banfield, 1955). For example, Hirsch (1983) argues that the politics surrounding the Chicago Housing Authority allowed individual communities to veto planned projects, leading to enormous concentrations of projects and their largely poor African American residents in just a few politically weak neighborhoods.

The Civil Rights Act of 1968 and several court decisions, including Gautreaux v. The Chicago Housing Authority (1967) and Shannon v. HUD (1970), however, forced a change in assisted housing policy. In the 1970s, HUD adopted a series of siting standards that were designed to expand assisted housing opportunities beyond areas of minority concentration. Vernarelli (1986) recounts and explains the evolution of siting criteria through the mid 1980s. He characterizes HUD's policies promoting fair housing as ambiguous and uncertain, evolving out of complex interactions among the legislative, executive, and judicial branches of the federal government. A time line of the major policy events from 1962 through 1984 is shown in Figure 1. Although there have been major policy events since 1984, including the Fair Housing Amendments Act of 1988, assisted housing siting criteria have largely remained the same (see Roisman, 1999a, for list of recent desegregation suits against HUD).

HUD siting regulations, in summary, seek to achieve "balance between its duty to create expanded housing opportunities for minorities and to direct a fair share of housing resources to minority areas to serve the families who voluntarily choose to live there" (Vernarelli, 1986, p. 231). Since the late 1970s, HUD has reviewed the proposed sites of new developments funded under its programs to ensure that they were not located in areas of minority concentration unless there were insufficient opportunities outside areas of minority concentrations or an "overriding need" for the housing to be built there. Not being a HUD program, however, the LIHTC program has not been subject to the same siting regulations. In fact, the U.S. Treasury Department, which administers what has become the largest subsidized housing program in the country, has no regulations that address the siting of developments in areas of minority concentration (Roisman, 1999b).

Research suggests that in more recent years, race may have played a less important role in the siting of assisted housing developments. Warren (1986) reports that there was some dispersal of assisted housing away from African American neighborhoods in Baltimore and St. Louis during the 1970s, but it is not clear if these results are applicable to the nation at large or to areas of Latino concentration. Newman and Schnare (1997) found that while public housing developments are located in neighborhoods with large minority populations, other types of assisted housing are sited in more integrated neighborhoods. Moreover, since most public housing units were sited prior to the enactment of the fair housing laws and siting criteria, it is not appropriate to conclude that the more recent sitings of public housing developments have favored minority areas. Finally, in an analysis of 10 cities, Gray and Tursky (1986) found assisted housing programs implemented after 1968 to be sited in less racially concentrated areas than those of the public housing program.

Enlarge 200% Enlarge 400%

Thus, although the evidence clearly illustrates that race has historically played a key role in determining the location of public housing, it is not clear whether these historic patterns have persisted into the 1980s. The policy changes and court orders of the 1960s and 1970s may have been effective in expanding the types of neighborhoods in which public and other assisted housing is located. It is also not clear whether assisted housing has been directed toward areas of Latino concentration or whether other types of assisted housing besides public housing have been targeted toward minority neighborhoods. Finally, the question of how large the impact of race is on the impacting of assisted housing remains unanswered by prior research.

The role of race and ethnicity may also vary depending on whether assisted housing developments are occupied by families or the elderly. The siting of elderly assisted housing may be less determined by neighborhood racial composition for three reasons. First, by regulation, the residents of elderly housing developments have to be at least 62 years of age. Given their ages, these residents are not likely to be seen as threatening. Second, although tenants of elderly assisted housing are poor, their poverty is less likely to be attributed to any personal defects, thus they are more likely to be seen as acceptable neighbors. Indeed, Newman and Schnare (1997) report that the fraction of assisted family units located in underclass neighborhoods is more than three times the number of assisted elderly or handicapped units. Finally, elderly housing is more likely to have White occupants (Goering, 1994; Goering et al., 1997), again making the tenants more acceptable to White neighbors.

The available evidence does show that public housing developments for the elderly are located in neighborhoods with larger proportions of Whites (Goering, 1994; Goering et al., 1997). The racial composition of neighborhoods does not appear to factor into siting decisions for elderly housing to the extent that it does for family housing. Consequently, the political economy of race model is likely to be a weaker predictor of the siting of elderly housing than the siting of family housing.

A second perspective that has been used to explain the siting of public and assisted housing is drawn from urban ecological theory. This perspective suggests that the position of a particular neighborhood vis-a-vis the overall metropolitan landscape is a key determinant in the siting of public and other forms of assisted housing (Taeuber & Taeuber, 1965; Goldstein & Yancey, 1986). Adherents to this school of thought argue that areas close to the central business district represent the least desirable residential spaces, and that desirability increases as one moves toward the urban periphery. As more affluent residents and businesses flee the central city for suburbia, properties in inner-city neighborhoods decline in value, and in some cases, are abandoned. This migration results in relatively cheap vacant land with access to infrastructure and public services. Such sites are attractive for the development of public and assisted housing, and it is in these neighborhoods that assisted housing is likely to be located.

Local government regulation supports this process by enacting land use regulations to preserve and enhance property values. Since property taxes are the main source of local government revenue, municipalities are highly motivated to protect land values and prevent cross-subsidization by forming economically homogeneous communities (Tiebout, 1956). By discouraging or preventing the construction of lower-cost homes in otherwise affluent areas, local regulations limit the supply of low-cost housing in the more desirable suburban areas (Schill & Wachter, 1995) and thus relegate the development of public and assisted housing to central-city areas where land is cheaper and the regulations more accommodating to multifamily housing.

The available evidence from case studies in Chicago and Philadelphia tends to support the political economy of race theory rather than urban ecological theory (Goldstein & Yancey, 1986; Massey & Kanaiaupuni, 1993), but it is uncertain whether these results are applicable to other cities, to other types of assisted housing, or to developments built in the 1980s, when HUD policies should have made race a lesser factor in the siting of assisted housing than in earlier times.

Consequently, although this investigation focuses on the role of neighborhood racial composition in the siting of assisted housing, the urban ecological model suggests other factors that must be controlled for in the analysis. Whereas the political economy of race school points to the racial composition of a neighborhood as the key determinant in the location of assisted housing, urban ecology suggests that older, less expensive neighborhoods located near the center of the city will be likely candidates for the development of assisted housing.

Study Methods

This study was part of a larger one that sought to determine what factors influenced the siting of assisted housing and what impacts the assisted housing developments had on the surrounding neighborhoods once they were built (Freeman & Rohe, 1998, 2000). This article represents the first stage of that research, examining the patterns behind the siting of assisted housing.

As mentioned above, most of the research on the siting and impacts of assisted housing developments has been confined to public housing. One of the goals of our research is to identify the neighborhood characteristics associated with the siting of other types of assisted housing. Thus, in addition to public housing developments, we also consider two additional types of assisted housing built during the 1980s: Low Income Housing Tax Credit developments and what we call other HUD projectbased subsidy programs, which include developments built under the Section 8 New Construction, FHA Multi-- family, and Section 236 programs.

As mentioned earlier, it is important to distinguish between assisted housing for families and for the elderly. To do this, we divide assisted housing developments into two groups depending on whether 50% or more of the household heads are over age 62. We do this for public housing and other HUD assisted housing separately.1 Unfortunately, no demographic information is available for LIHTC developments, so we cannot distinguish predominantly elderly or family LIHTC developments. Thus, the analysis will focus on five discrete types of assisted housing developments: family public, elderly public, other HUD family, other HUD elderly, and LIHTC housing developments.

Our approach to the analysis begins with the presentation of data on the likelihood that each of the five types of assisted housing was sited in neighborhoods with particular social and physical characteristics. For example, we calculate the relative probabilities that census tracts with different percentages of African American residents received family public housing during the 1980s. After considering the patterns found in this bivariate analysis, we specify logistic regression models for each type of housing to assess the combined effects of tract social and physical characteristics on the siting of the different types of assisted housing developments.

Our choice of independent variables used to model the likelihood of a neighborhood receiving assisted housing was guided by the political economy of race and urban ecology theories. As noted above, the political economy of race theory posits that politically impotent neighborhoods will receive a disproportionate share of undesirable land uses, such as assisted housing. In urban America, politically impotent neighborhoods have typically been synonymous with poor, minority communities. To capture the effect of a neighborhood's racial composition on the likelihood of receiving assisted housing, we include in our analysis the proportion of neighborhood population that is African American, the proportion that is Latino, and the proportion that is Asian. Because African Americans and Latinos are typically thought of as two of the most disadvantaged groups in America, we suspected that neighborhoods with high concentrations of both of these groups might be especially susceptible to receiving unwanted land uses such as assisted housing. To test this hypothesis, we include an interaction term that is the product of the percent Latino population and the percent Hispanic population in a tract.2 We reason that areas with a mix of minority groups may be fractured politically and this may hamper their abilities to influence the siting process.

We also include measures of the neighborhood's socioeconomic status in our analysis, including neighborhood poverty rate, the price of owner-occupied housing in a tract as a percentage of the median house price in the metropolitan area, and household income in a tract as a percentage of the median household income in the metropolitan area. The proportion of housing in a neighborhood that is single family and owner occupied is also included. We reason that homeowners might be especially resistant to the notion of assisted housing developments being built in their neighborhoods because of fear that they will lower the values of surrounding houses (Freeman & Rohe, 2000). In addition, a neighborhood consisting mostly of single-family, owner-occupied units might be less likely to have the infrastructure or zoning that would allow the development of multifamily units that are typical of assisted housing. Thus, the proportion of single-family, owner-occupied housing units in a neighborhood, in addition to being a proxy for neighborhood resistance to assisted housing, is an ecological factor that might influence whether assisted housing is built in a neighborhood.

The urban ecology school also suggests that older neighborhoods and those closer to the core of the metropolitan area will have more land that is no longer viewed as desirable by private developers, and, consequently, is less expensive. To capture a neighborhood's ecological characteristics, we include the distance between the centroid of the neighborhood and the centroid of the metropolitan area's central business district (CBD), the proportions of housing in the neighborhood built before 1950 and 1970, and the proportion of units that are vacant. The neighborhood's distance from the CBD was coded as five dummy variables representing a distance of zero to 2.5 miles, 2.51 to 5 miles, 5.1 to 10 miles, 11 to 25 miles, and greater than 25 miles.

In addition to the factors described above, we also control for whether the neighborhood is in the Northeast, Midwest, South, or West; the proportion of a neighborhood's population at age 65 years or older; and the proportion of the area's housing that was public and assisted prior to 1980. We include a measure of the concentration of elderly in a neighborhood because the elderly might be especially active in fighting family-- occupied assisted housing, yet be receptive to elderly-- occupied assisted housing. We include the proportion of assisted housing prior to 1980 to discern if there has been any effort to disperse assisted housing away from existing concentrations.3

To determine the location of assisted housing, we used data from HUD's 1997 Picture of Subsidized Housing (PSH), which contains information on public housing, Section 236, Section 8 New Construction, FHA, and LIHTC developments. HUD provided us with construction completion dates for each development in a separate data set. The 1990 census tract of each development is contained in the PSH, along with demographic information on the occupants of assisted housing developments as of 1997. We used the tract identification number to link 1980 tracts to 1990 tracts in all metropolitan areas in the country. Because we are using the characteristics of neighborhoods in 1980 (from the STF3 Census file) to predict the siting of assisted housing, we had to exclude from our analysis developments in areas that were tracted only after 1980. This amounted to 5,170 of the tracts in 1990, leaving 39,779 tracts used in our analysis of the siting of assisted housing. Thus, in the analysis presented below, 1980 census tract data is used to predict whether the tracts in our sample did or did not receive the different types of assisted housing during the 1980s. Due to the dichotomous nature of the dependent variable, logistic regression is used in the multivariate analysis.

Findings

Bivariate Analysis

We begin by presenting the bivariate data on the relationship between the key neighborhood characteristics considered and the siting of each type of assisted housing.4 The bivariate analysis involves calculating probabilities that indicate the likelihood that census tracts with certain characteristics received a particular type of public and assisted housing compared to the overall probability of tracts receiving that type of assisted housing. This is calculated by dividing the percentage of tracts in each category of the independent variable by the percentage of all tracts that received that type of assisted housing development. For example, 2.23% of all tracts in our sample received family public housing during the study period. Yet, only 1.57% of the tracts with less than 5% African American population received family public housing, a ratio of .70, while 3.19% of tracts with more than 50% African American population received family public housing, a ratio of 1.75. Ratios less than 1 indicate that tracts in a category are less likely than would be expected, given an even distribution among categories, to receive a particular type of assisted housing, while ratios over 1 indicate a greater than expected percentage of assisted housing developments.

Of particular interest to this study is the siting pattern for tracts with various racial and ethnic characteristics. The results show that the probabilities of family public housing, other HUD family housing, and LIHTC housing developments being sited in census tracts increased with the percentage of African American households in those tracts (see Table 1). This pattern is most pronounced for the LIHTC developments. Tracts with a majority of African American households were 2.45 times as likely to receive a LIHTC development, while these tracts were 1.82 times as likely to receive other HUD family housing and 1.75 times as likely to receive family public housing. Neither of the two types of elderly housing developments, however, were concentrated in predominantly African American tracts.

A similar, although not quite as pronounced, pattern was found between the siting of the different types of assisted housing and the percentage of Hispanic population (see Table 1). Family public housing was 1.90 times as likely, other HUD family housing was 1.68 times as likely, and LIHTC developments were 1.34 times as likely to be sited in tracts with a majority of Hispanics. Elderly public housing was also more likely to be sited in predominantly Hispanic areas, but there was no such concentration for other HUD elderly housing.

The siting pattern with respect to the percentage of Asians in tracts is quite different. Elderly public housing was less likely to be sited in areas where Asians made up more than 20% of the households, while other HUD family and elderly housing developments were more likely to be built in areas with more than 20% Asian households (see Table 1).

The combined percentage of African American and Hispanic households is also associated with the siting of all five types of assisted housing, although the family developments show the strongest associations (see Table 1). LIHTC developments were particularly likely to be sited in areas with a mix of African American and Hispanic households.

We also wanted to see how the siting of public and assisted housing developments varied by the percentage of poor households, the median household income, and the percentage of elderly in tracts. The data indicate that four of the five types of subsidized housing were more likely to be sited in tracts with higher poverty rates (see Table 2). This pattern was particularly evident for both types of family housing and for LIHTC developments. Each of these three housing types was more than twice as likely to be sited in areas where more than 40% of the households were below the poverty line. LIHTC developments were particularly likely to be sited in tracts with households in poverty exceeding 25%.

Enlarge 200% Enlarge 400%

All types of assisted housing were also more likely to be sited in tracts with relatively low household incomes (see Table 2). With the exception of elderly public housing developments, the likelihood of an assisted housing development being sited in a very-low-income tract was twice the expected rate. It is interesting to note that the modest-income tracts were the least likely to receive a family public housing or a LIHTC development.

Turning to the siting of subsidized housing developments in tracts with differing physical and housing value characteristics, the data indicate that tracts with relatively lower housing values were much more likely to receive all types of assisted housing (see Table 3). This pattern was particularly pronounced, however, for LIHTC, other HUD family, and family public housing developments. A tract in the lowest value category had over 14 times the probability of receiving a LIHTC development than a tract in the highest value category. The percentage of single-family, owner-occupied units in tracts is also related to the likelihood of receiving each type of assisted housing. The pattern is consistent for all housing types: tracts with higher percentages of single-- family, owner-occupied units were consistently less likely to receive assisted housing developments (see Table 3).

The data on the probability of placing subsidized housing in tracts that already contained public housing indicates that all assisted housing types were more likely to be sited in tracts with existing public housing (see Table 3). Tracts that already had public housing, for example, were more than three times as likely as the average tract to receive a new public housing development and more than four times as likely to receive an elderly public housing development.

Enlarge 200% Enlarge 400%

Enlarge 200% Enlarge 400%

Multivariate Analysis

To assess the relative importance of each independent variable after controlling for all other observed variables, we specified five logistic regression models, one for each type of housing. The standardized regression estimates for each of these models are presented in Table 4.

A central question of this research was whether race continued to play a dominant role in determining where assisted housing was located during the 1980s. The results suggest that once the influences of other variables were controlled, the percentage of African American residents was not a significant factor in explaining the siting of public or other HUD family housing, nor either type of elderly housing. The percentage of African American residents did, however, show a statistically significant association with the siting of LIHTC developments. The standardized regression coefficient for LIHTC developments was a relatively strong .127.

This research also sought to assess the role of ethnicity in the siting of assisted housing. The results of the logistic regression models indicate that once the influences of other variables were controlled, the percentage of Hispanics in a census tract did not show a statistically significant association with the siting of any type of assisted housing. The standardized regression coefficients in each of the five models are all under .04. The percentage of Asians in census tracts, however, is negatively related to the probability of receiving two types of assisted housing; both family and elderly public housing are less likely to be built in areas with higher proportions of Asians. The standardized coefficients are -.106 and-.186 respectively. The reason for this may have to do with the lack of sites suitable for public housing in Asian areas, which, because of rapid growth, may have less vacant land suitable for new assisted housing developments.

These findings suggest that race and Hispanic ethnicity did not play a significant role in the siting of assisted housing during the 1980s. The variable representing the product of African American and Hispanic households, however, indicates that they did. Neighborhoods with relatively high proportions of both African American and Hispanic households were more likely to receive both family public and other HUD family housing developments. The standardized coefficient for family public housing is .046, while that for other HUD housing is .031. In addition, the direction of the impact for the percentage of African Americans variable is positive for the placement of family assisted housing and negative for the placement of elderly assisted housing, contrary to what one would expect if race did not matter. The findings concerning the relationship between the income characteristics of tracts and the placement of assisted housing indicate that four of the five types of housing are more likely to be built in areas with higher percentages of persons in poverty. The exception is elderly public housing, which is no more likely to be built in higher poverty areas. The standardized coefficients range from .063 for LIHTC developments to .132 for family public housing developments. Once the poverty population is accounted for, the median household incomes of tracts is only weakly related to the receipt of subsidized housing developments. The findings also indicate that elderly public, elderly subsidized, and LIHTC housing were more likely to be built in areas with higher percentages of persons over the age of 65. Other HUD family housing, however, was less likely to be sited in tracts with relatively high percentages of older residents.

The results of the logistic regression models indicate that for four out of the five types of assisted housing, the relative price of owner-occupied housing in tracts was a significant predictor of the location of assisted housing developments. The standardized regression coefficients for the family and elderly public, other HUD family, and LIHTC developments are all strongly negative, indicating that lower property values are strong predictors of the siting of these types of housing. The exception to this pattern was other HUD elderly developments. The standardized coefficient for family public and LIHTC housing are particularly strong at -.314 and -.466 respectively.

The percentage of single-family homeowners in tracts is significantly related to the siting of four of the five types of assisted housing developments studied. The coefficient for elderly public, other HUD family, and other HUD elderly developments are all negative, indicating that those developments were less likely to be placed in areas with relatively high percentages of single-- family, owner-occupied housing. The coefficient for LIHTC developments is positive, however, indicating some tendency for those developments to have been sited in areas with higher percentages of single-family, owner-- occupied units. The percentage of vacant units in tracts was negatively related to other HUD family housing and positively related to the siting of LIHTC developments.

The relationships between the age of housing in tracts and the siting of assisted housing developments show interesting variation among the types of assisted housing analyzed. Other HUD family housing developments were less likely to be located in areas with older housing (b = -.047), while LIHTC developments were more likely to have been sited in areas with older housing (b =. 146). The LIHTC finding is likely due to the greater use of the LIHTC program in funding the rehabilitation of existing structures.

The results of the logistic regression models also indicate that there is some tendency for housing developments built under particular programs to be placed in tracts with other HUD assisted units. All types of housing developments, with the exception of elderly public housing, were more likely to be sited in tracts with existing subsidized units, and other HUD elderly developments were more likely to be sited in tracts containing public housing developments.

The distance of census tracts from their respective central business districts (CBDs) is also related to the probability of tracts receiving several types of assisted housing. With the exception of the first distance category, which for many cities will contain large proportions of commercial property, family public housing and other HUD family housing developments were most likely to be built in tracts closer to CBDs. As the distance of a tract from its CBD increased, the likelihood of a tract receiving these types of housing developments decreased. The standardized coefficients for other HUD elderly housing and LIHTC developments indicate that these developments were more likely to be sited in tracts further from the CBD.

Finally, the standardized logistic coefficients for the three dummy variables for region (the South as the excluded category) indicate that census tracts in the West were more likely to receive both family public housing and LIHTC developments. Census tracts in the Northeast were less likely to receive LIHTC developments. Finally, tracts in the Midwest were more likely to receive both other HUD elderly housing and LIHTC developments.

Enlarge 200% Enlarge 400%

Discussion

Two theoretical perspectives have been offered to explain the siting of assisted housing: the political economy perspective and the urban ecology perspective. The political economy of race perspective suggests that the siting of assisted housing is explained by the relative political power of subgroups in our society. Because they are often viewed as undesirable additions to a neighborhood, assisted housing developments are placed in areas inhabited by the politically weakest groups in society, areas with relatively large percentages of racial and ethnic minorities and the poor. The urban ecology perspective, on the other hand, suggests that assisted housing will be placed on relatively inexpensive sites, with existing public infrastructure close to CBDs.

The bivariate analysis presented here provides considerable support for both perspectives. Consistent with the political economy of race perspective, most types of assisted housing, particularly the family-oriented developments, were more likely to be developed in tracts with relatively high percentages of African American, Hispanic, or African American and Hispanic households. Moreover, they were also more likely to be sited in tracts with relatively high percentages of poor households. On the other hand, consistent with the urban ecology perspective, assisted housing developments were also more likely to be built in areas with lower housing values; lower percentages of single-family, owner-occupied units; and existing HUD assisted housing. Moreover, both types of HUD family housing developments were more likely to be sited in areas that are closer to the CBD.

The results of the multivariate analysis also suggest that both perspectives are important in fully explaining the siting of public and assisted housing developments. The variable with the strongest relationships to the placement of most types of assisted housing is the value of owner-occupied units in tracts, a variable associated with the urban ecology perspective. This variable is the strongest predictor of the placement ofboth family public housing and LIHTC developments and the second strongest predictor of the siting of elderly public housing. Other HUD elderly developments was the only type of housing not significantly related to the value of owner-occupied housing in tracts. Furthermore, the percentage of single-family, owner-occupied units in a tract, another urban ecology variable, was also a relatively strong predictor for the siting of four of the five types of assisted housing considered.

After controlling for the variables associated with the urban ecology perspective, however, race, ethnicity, and income still mattered. The percentage of African Americans in a neighborhood was a relatively strong predictor of the siting of LIHTC developments. Moreover, the combination of African American and Hispanic households in tracts was a predictor of the siting of both public family and other HUD family developments. This finding is consistent with the idea that mixed African American and Hispanic areas are more likely to be politically fractured and less able to influence the siting process.

Moreover, whereas the percentage African American population in a neighborhood had a negative impact on the likelihood of elderly housing being built there, the direction of the association was positive for family assisted housing. Although these impacts were not significant in the multivariate models, they do suggest that the impact of race on the siting of assisted housing is different for family housing than for elderly housing. Such a difference is inconsistent with urban ecology theory, but wholly consistent with the predictions of the political economy of race model. Overall, we conclude that despite HUD's efforts to deconcentrate assisted housing and numerous lawsuits with the same intention, race and ethnicity still play roles, albeit modest ones, in the siting of family-oriented assisted housing.

The multivariate results also show that four of the five types of assisted housing were more likely to be placed in tracts with relatively high proportions of poor households. This was particularly the case with family public housing developments. Given recent evidence on the negative effects of growing up in high-poverty areas, this is a particularly important finding (Brooke-Gunn et al., 1999; Duncan & Brooke-Gunn, 1997). It suggests that children growing up in assisted housing are not being afforded the same opportunity for success as other children.

Turning to policy considerations, one of the most damning criticisms of assisted housing, and especially of public housing, is that it has contributed to the "American apartheid" that characterizes many of America's urban areas (Massey & Denton, 1993). By targeting minority neighborhoods and spurring neighborhood racial transition, assisted housing served as an institutionalized mechanism for creating and perpetuating residential segregation. The question we attempted to answer here was whether the historic link between areas of minority concentration and the siting of assisted housing was broken in the 1980s.

Our results show that LIHTC developments were the most likely to be sited in areas with higher percentages of African American residents. This should not be terribly surprising for two reasons. First, not being a HUD-managed program, the LIHTC program has not been subject to the same siting regulations as the other development types. The siting standards under the LIHTC program are left to the individual state agencies that are responsible for administering the program, and many do not have provisions covering the racial makeup of the areas in which the developments are sited. Second, LIHTC developments in "qualified census tracts," which include tracts in which 50% or more of the households have incomes below 60% of area median income, are eligible for larger tax credits than usual. Given the lower average incomes of African Americans, those areas are more likely to contain a higher percentage of African Americans. Thus, LIHTC regulations actually favor the siting of developments in locations more likely to be areas of minority concentration.

The lack of attention given to the issue of the concentrations of LIHTC developments in minority neighborhoods is a serious gap in the federal effort to guard against housing segregation. There is no logical reason why the siting of LIHTC developments should not receive the same scrutiny as other forms of assisted housing. (See Roisman, 1999b, for a proposal to amend the LIHTC regulations.) Aside from the LIHTC program, our results suggest that the regulations designed to stop the concentration of assisted housing developments in predominantly African American areas seem to have lessened the role of race in the siting of assisting housing, although race has not been completely eradicated as a determinant of where some types of assisted housing is built.

The important policy question is whether the degree of concentration in areas with higher proportions of African American and Hispanic residents is enough to warrant additional actions to discourage the development of assisted housing developments in those areas. In our opinion it is not, because our analysis indicates that during the 1980s many developments of all types were built in areas with high percentages of White residents. Thus, the housing developments built during the decade did provide many opportunities for assisted housing residents to live in areas with low percentages of African American and Hispanic residents. This is in stark contrast to the siting of assisted housing in post-World War II Chicago, where, as portrayed by Hirsch (1983) and Myerson and Ban field (1955), virtually no public housing was built in White neighborhoods.

It is important to remember that the siting requirements were never meant to prohibit development in areas with high percentages of African Americans; they were only intended to provide a wider range of choice for public and assisted housing residents. It is not reasonable to expect that assisted housing developments will be evenly distributed across tracts that vary in land value and access to services such as public transportation.

Moreover, we certainly would not advocate prohibiting the development of assisted housing developments in African American and Hispanic neighborhoods. This is particularly true now that these developments tend to be better designed, smaller in size, and more diverse in the income levels of residents than the ones built before 1980. These developments may play an important role in revitalizing these areas and providing local residents with better, more affordable housing. Such a strict policy would likely have a negative impact on the missions and financial support of community development corporations, which are often involved in developing LIHTC housing in areas with a relatively high proportion of minority residents.?

Additional analysis not presented in this paper does indicate, however, that there is great variation among metropolitan areas in the degree to which assisted housing developments were sited in areas with high proportions of African American residents (Freeman & Rohe, 1998). In the Baltimore and St. Louis metropolitan areas, for example, large percentages of most types of assisted housing developments were sited in areas of minority concentration. A likely explanation for this concentration is that their central cities, which have relatively high percentages of African Americans, are actively pursing new assisted housing developments, while the surrounding suburban jurisdictions, which have a relatively high percentage of White residents, are not.

In Dallas, however, large percentages of most types of assisted housing were sited in areas with relatively low percentages of African Americans. To some extent this may be due to the relatively large proportion of Hispanics in the Dallas area, but the relative lack of governmental fragmentation in the metropolitan area is likely to contribute to this pattern. Federal policy might address this problem by encouraging, through both incentives and sanctions, the development of assisted housing in areas that have shunned it in the past. Such a policy would do a lot more to open up additional opportunities for residents of assisted housing than tightening the siting criteria for communities that are working to increase the supply of assisted housing in their jurisdictions.

In conclusion, the analysis in this article supports the notion that HUD policies designed to provide residents of assisted housing developments with more choice in terms of the racial composition of the neighborhood in which they reside were somewhat successful. Although there is still a statistically significant relationship between the racial and/or ethnic characteristics of neighborhoods and the placement of some types of assisted housing, these relationships are generally weak. Unlike the pre-1970s pattern of placing a large majority of family assisted housing developments in African American neighborhoods, in the 1980s many of these developments were built in predominantly White areas.

ACKNOWLEDGMENTS

We would like to acknowledge and thank several persons who provided valuable assistance with this research project. Shannon Van Zandt, a Ph.D. student at UNC-Chapel Hill, assisted with the literature review. Miki Satake, at Mathamatica Research, did much of the programming required to merge the data sets used in this study. Their contributions were critical to the completion of this study. The work that provided the basis for this article was supported by grant funding from the U.S. Department of Housing and Urban Development. The substance and findings of the work are dedicated to the public. The authors are solely responsible for the accuracy of the statements and interpretations contained in this publication. Such interpretations do not necessarily reflect the views of the government.

Abt Associates. (1996, July 1). Development and analysis ofthe national low-income housing tax credit data base: Final report, Vols. 1 &2. Boston: Author.

Bauman,J. F. (1987). Public housing, race, and renewal: Urban planning in Philadelphia, 1920-1974. Philadelphia: Temple University Press.

Brook-Gunn, J., Duncan, G., & Abler, L. (Eds.). (1999). Neighborhood poverty: Context and consequences for children. New York: Russell Sage Foundation.

Civil Rights Act of 1968, 42 U.S.C. 3600-20 (1968).

Duncan, G., & Brooke-Gunn, J. (Eds.). (1997). Consequences of growing up poor. New York: Russell Sage Foundation.

Fair Housing Amendments Act of 1988, 42 U.S.C. 3601 et. seq. (1988).

Freeman, L., & Rohe, W. (1998). The siting of assisted housing and its impact on neighborhood racial transition. Chapel Hill: Center for Urban and Regional Studies, University of North Carolina at Chapel Hill.

Freeman, L., & Rohe, W. (2000). Subsidized housing and neighborhood racial transition: An empirical investigation. Housing Policy Debate, 11 (1), 67-89.

Galster, G. (1999). The evolving challenges of fair housing since 1968: Open housing, integration, and the reduction of ghettoization. Cityscape, 4(3), 123-138.

Galster, G., & Keeney, H. (1993). Subsidized housing and racial change in Yonkers, New York.Journal of the American Planning Association, 59, 172-181.

Gautreaux v. The Chicago Housing Authority et al., 265 F. Supp. 582 (N.D. Ill. 1967).

Goering, J. (1994). The location and racial composition of public housing in the United States. Washington, DC: U.S. Department of Housing and Urban Development, Office of Policy Development and Research.

Goering, J., Kamely, A., & Richardson, T. (1997). Recent research on racial segregation and poverty concentration in public housing in the United States. Urban Affairs Review, 32(5), 723-745.

Goldstein, L, & Yancey, W. L. (1986). Public housing projects, blacks, and public policy: The historical ecology of public housing in Philadelphia. In J. M. Goering (Ed.), Housing desegregation and federal policy (pp. 262-289). Chapel Hill: University of North Carolina Press.

Gray, IR, & Tursky, S. (1986). Location and racial/ethnic occupancy patterns for HUD-subsidized family housing in ten metropolitan areas. In J. M. Goering (Ed.), Housing desegregation and federal policy (pp. 235-252). Chapel Hill: University of North Carolina Press.

Hirsch, A. (1983). Making the second ghetto: Race and housing in Chicago, 1940-1960. New York: Cambridge University Press.

[Author Affiliation]

Rohe is a professor of city and regional planning and the director ofthe Center for Urban and Regional Studies at The University of North Carolina at Chapel Hill. He is co-author of Planning with Neighborhoods (University of North Carolina Press, 1985) and more than 40 journal articles on housing and community development policy and practice. Freeman is an assistant professor in the Urban Planning Program at Columbia University. He has worked as a city planner for the NewYork City Housing Authority and has a Master's and a Ph.D. in city and regional planning from The University of North Carolina at Chapel Hill.