Roy Elliot Cundiff GRG 394k Dr. Hudson Spring 1999
Introduction
Residential segregation involves the spatial concentration of population groups. This phenomena is not new; various societies have segregated the inhabitants of their settlement space according to a variety of criterion for thousands of years. The most common forms of segregation are economic segregation and segregation along racial or ethnic lines. In the United States, and for the purposes of this paper, residential segregation will be defined as the tendency for individuals with different racial backgrounds to inhabit different parts of metropolitan areas in greater concentrations. In the United States, segregation is often associated with the income level of the racial or ethnic group in question. Consequently, the following discussion of residential segregation will involve complicated relationships between economic and racial segregation. It will, like other such studies, involve the application of quantitative methods to census data in an effort to explain situations that deal with human attitudes, biases, and preferences. This can be problematic, and therefore the results offered here will be explained in light of the limitations of such methods.Researching American Segregation
The study of residential phenomena is not very old, with much pioneering work occurring at the University of Chicago in the 1920's and 1930's. The "Chicago School" of urban scholars in sociology and urban geography were reform minded in their study of segregation. Much academic study in urban settings addressed problems associated with tenement housing, the externalities associated with proximity to industry, and crime and gang violence which tended to concentrate in certain areas of a given city. Much of the work of Homer Hoyt, Louis Wirth, Robert Park, and others concerned the concentration of societal problems in areas of Chicago that were inhabited largely by immigrants from foreign countries and African Americans from the American South (Gottdeiner 1985). As such, the early study of segregation constituted an attempt to redress social inequalities across racial and ethnic lines. This purposeful, goal-oriented approach was consistent with the "preoccupation in society at large that newcomers needed to be integrated into society" (Van Kempen and Ozuekren 1998: 1632). This was certainly the case with much of the segregation literature during the Civil Rights movement and the de jure desegregation of schools and other public facilities in the 1950's and 1960's.
In recent years, academic research has often not attempted to direct policy, and is instead more concerned with an accurate, methodical documentation of the state of segregation and its causes, along with possible future trends in an ever more pluralistic nation. Growing solidarity and political organization amongst racial, ethnic, and neighborhood groups has changed the discourse of race, segregation, and power. Now, it is far less common for academics to prescribe solutions based on the relocation of populations over space. By joining together into powerful political voting blocks and neighborhood organizations, many racial groups can makes demands for social services that were denied to them in the past. There is some concern that this power in diminished when racial or ethnic groups integrate into the larger society, and attempts to "desegregate" some areas are often now associated with the negative aspects of gentrification. This process involves the restoration and investment in certain neighborhoods, which often leads to increase property values, an influx of wealthy white professionals, and displacement of original populations who can no longer afford to pay taxes or rent.Housing Stock, Integration, and the Tipping Point
The spatial structure of the American city is often dictated by zoning and residential neighborhoods that are similar in terms of the style, size, and age of the housing stock. These areas will consequently become associated with residents of certain income levels that can more readily afford to live there. The result is often whole sections of a city that have long been "wealthy, white neighborhoods" or "poor black neighborhoods". In older industrial American cities of the Northeast and Midwest, the result is often known as "overstructuring", or "ghettoization", whereby obsolete housing becomes associated with a specific racial group over time, decreasing the chances of racial change in that area (Hartshorn 1992: 269, Datel and Dingemanns 1995: 458-9). The presence of a "dual-market" for white and black residents of certain cities, sometimes accompanied by unscrupulous real estate practices, has reinforced segregation in some areas. In the South, the existence of legal segregation has also resulted in the association of certain neighborhoods or entire sections of cities with members of a certain race. These associations in conjunction with racial attitudes can make change difficult in many of these areas, which are likely to remain similar for many generations.
A small body of literature has dealt with the concept of integration, and how best to define this phenomenon. Smith (1998) describes the two streams of thought on this subject as "demographic integration" and "social integration". The latter implies a societal cooperation between members of two groups under conditions of equality, and is best studied in small scale neighborhood studies with qualitative research. Demographic integration is defined as the mixing of races within urban space; this is the method commonly undertaken with the tools and methods of demographic research: census data, quantitative analysis of trends, etc. Definitions vary as to what constitutes a truly integrated neighborhood, because some researcher have used a 50/50 split as a benchmark, while others have protested that this may not be a good indicator in a country that is 75% white, 12% black, roughly 9% Hispanic, and 3% Asian (Ellen 1998: 28). Also, any census data will likely give the analyst a snapshot frozen in time. A neighborhood that appears "integrated" may only be in transition. Pessimists believe that in some place integration is the time between the first non-white resident arriving and the last white resident departing. This is consistent with what is known as the "tipping point" theory (Hartshorn 1992: 294). The tipping point is the highest number of minority households that will be tolerated in a formerly white neighborhood before white residents begin to move. This can result in a chain reaction of "white flight", often due to an unsubstantiated fear that property values will decrease. Because of this phenomena, it is more difficult to categorize integration without detailed time series data. This can be difficult to work with due to frequently changing racial categories and census boundaries.Racial Isolation
Recent studies by Massey and Denton (1989, 1993) and others have demonstrated decreasing levels of residential segregation for metropolitan areas. Minority populations accounted for 18% of the nationwide suburban population in 1990, as opposed to 13% in 1980 (O'Hare 1993). Massey and Denton also noted the presence of "hypersegregation", their term for the presence of extremely concentrated inner city clusters of African Americans, even while metropolitan area segregation levels are seen (and quoted) to be falling. Segregation literature accounts for these seemingly mutually exclusive trends by pointing out that even small minority increases in the suburbs lowers measured levels of metropolitan segregation. However, these new suburbanites are not necessarily replaced by inner city white residents. The result is suburbs that are slowly becoming more diverse as minority groups with financial means move into the periphery (or to areas of the city that have historically been primarily white). Meanwhile, certain sections of the inner city remain "hypersegregated" and racially homogenous.
These inner city areas are often more than 90% black in racial composition, and large parts of many suburbs are often over 90% white. Wilson's concern is that this process concentrates inner city poverty and social problems in certain areas, leaving "disadvantaged" minorities behind in places with few role models, poor employment training opportunities, and little chance for advancement through education (Wilson 1991:8). This early development of "weak attachments to the labour force" reinforces inequalities of income according to race, in a cycle that is difficult to change (Carter et al 1998:1897).
It is largely thought that educational opportunities result in proportional increases in salaries and an ability to purchase housing in areas that were formerly closed to non-white groups. In the case of African-Americans (the term "black" will also be used extensively in this paper because it is the official racial category used by the U.S. Bureau of the Census), education and earning power have been a major hindrance to neighborhood integration (Wilson 1987).
It seems that the recent focus on inequalities between racial groups has sought to address solutions based on training, education, and social services to areas of the city that need them. In a "color-blind", equitable society, with equal opportunities for all of its citizens, individuals from any background should have equal access to choices in housing, employment, and education. Unfortunately, an equal society may not exist. American society contains racial inequalities that are not easily redressed, and the constantly changing racial composition of the nation will create unknown changes in future patterns of residential segregation. In their study of the complicated patterns emerging in Los Angeles county in recent decades, one group of researcher coined the term "prismatic metropolis" (Zubrinsky and Bobo 1996: 336).Census Data
There are some major problems with the use of Census data in the study of residential segregation. The four major racial groups used by the 1990 Census are White, Black, American Indian, Asian/Pacific Islander, and Other. Hispanic is NOT a racial category in the Census, but rather an additional question of ethnicity above and beyond race. This means that Hispanic respondents can be white, black, or "other" in race. 98% of the 1990 Census respondents who checked "other" in the category of race also marked themselves as Hispanic. The rest of the "Other" category are people of mixed racial background. For ease of manipulation, most downloadable Census data has been pre-corrected to include Hispanic as a separate racial group. In our multicultural society, this problem will only get worse as mixed marriages break down the increasingly arbitrary categories we have in place for race. Efforts that are being made to increase accuracy of census data by including more categories have met with opposition due to fear that the surveys will become too difficult for people to fill out:As a result, all Census data must be accepted with some trepidation. While some researchers call for an end to racial categories all together, claiming that the existence of the categories may perpetuate difference amongst them, administrative functions require racial breakdown to insure fairness in housing lending, voting rights, and other programs that are based on equitable treatment across racial boundaries.Unfortunately for survey designers, the race problem can't be solved simply by adding new categories for mestizos and multiracials. The reasons are a pair of statistical bugaboos called "primacy" and "recency" effects. Studies show that when respondents see a long list of choices on a written survey, they are likely to pick the first choice that applies to them instead of reading the whole list. The reverse is true of telephone surveys, when people tend to pick the last choice. The more categories on the list, the more these effects complicate the results (Sandor 1994: 3).GIS and Segregation Indices
The use of Geographic Information Systems in the study of residential segregation is not very common. Much of the existing literature on the subject of residential segregation is based upon quantitative methods that do not use or even mention the use of G.I.S. Many of the measures used to describe segregation levels are not compatible with G.I.S. technology. Some of the more recent literature concerning G.I.S. and segregation has come from Wong (1993, 1997, 1998) and Anselin (1998), both of whom discuss segregation as secondary to other purposes (Wong's primary purpose seems to be the development of formulas and techniques, while Anselin seems primarily to be interested in the application of spatial statistics in real estate to the G.I.S. environment). Both of these authors have developed powerful statistical methods to measure segregation which require ArcInfo and statistical software such as SpaceStat or S-Plus (which features a "GIS-Link" to ArcInfo). The dearth of literature involving the specific application of G.I.S. to studies of segregation seem to be related to the following factors:-the problem of adjusting existing segregation measures and spatial statistics to the GIS environment
-preference of researchers for computer science aspects, statistical aspects, or social science aspects of either GIS or segregation (rarely do all of these interests overlap)Another hindrance is the problem of "spatial autocorrelation", often ignored in statistical analysis of geographic phenomena. Spatial autocorrelation, in short, deals with the tendency for geographic phenomena to cluster or occur in close proximity (Ebdon 1985: 150, Anselin 1998). In other words, successive values along a regression line may be strongly related to one another, varying in a systematic way. This is at odds with statistical measures that assume randomness of a sample- location is a variable in and of itself that has great predictive value and must be incorporated into statistical indices. Only recently has this been taken into account in segregation research (Wong 1993, 1997, 1998; Morrill 1991).
Overview of Project
The following report will utilize ArcView G.I.S. (versions 3.0a and 3.1) with SPSS (Windows v. 9.0) statistical software for the purposes of measuring levels of residential segregation in select Standard Metropolitan Statistical Areas (SMSA's). The purpose of the project is as follows:to identify various means of measuring residential segregation
to use the capabilities of Geographic Information Systems and statistical software to analyze Census data
to gain insight into the patterns of residential segregation in certain metropolitan areas
to assess the utility of GIS for this type of analysis
to identify future avenues of research into segregation, specifically involving G.I.S.Methodology
The most popular and practical means of measuring residential segregation (and most other demographic phenomena) is the use of census data. Census data is freely available and can be downloaded in digital format. Additionally, line files for use with Geographic Information Systems can also be downloaded from various sources. This project will use 1990 U.S. Census data at block group level, in conjunction with U.S. Census TIGER 95 boundary file data (an improvement over 1992 TIGER lines). Block group data, though more difficult to work with than tract level data because of the sheer number of areal units involved, provides greater accuracy. Census tracts, particularly in outlying regions of Metropolitan areas, can be enormous and all data for those areas will therefore be highly generalized. The project will consist of the following steps:-Choice of Metropolitan Areas
-Acquisition of data
-Preliminary analysis of results in ArcView, comparison between MSA's
-Importation and statistical analysis of census data and index measurements within SPSS
-Mapping of results in ArcView
-Correlation analysis in SPSS
-Conclusions and Suggestions for Future ResearchThe current project was limited by the necessity of using ArcView and SPSS, and unfortunately the modified segregation indices devised by Wong and Morrill could not be adapted to the ArcView environment. Instead, the standard (D) segregation index was used, along with display and calculation of black racial isolation in each study area. The link between these software packages is limited to the export of files from SPSS to a format (such as DBF) that is readable as an import file to ArcView. Because of this limitation the following report will consist of ArcView and SPSS components. The two measurements of segregation decided upon were the index of dissimilarity or segregation index (D) originally used by Duncan and Duncan (1955), and the measure of racial isolation or homogeneity. The segregation index is more readily applied in the ArcView GIS environment than modified indices that incorporate adjancy information such as binary contiguity matrices. This index is as follows:
D = (Sigma g1i/G1- Sigma g2i/G2)*100 (Duncan and Duncan 1955)
g1i, g2i = residents of racial categories 1 and 2 in areal unit i
G1 and G2 are the total populations for each group in the larger study areaThis index is applied by identifying all of the census block groups in a particular study area (in this case, an MSA), and totalling the population in these areas. The segregation index measures the dissimilarity between two groups. For the purposes of this project, black-white segregation will be measured.
A census block group is considered to be racially homogenous if 90% of its residents are of the same racial group. These phenomena will be mapped for each MSA and displayed in ArcView. The correlation between race and income will then be measured for each city to check for variation. It is assumed that variations will be slight, but that there will be relatively strong correlations between the number of white residents in a block group and the median household income of that block group. It is also assumed that there will be a negative correlation between the percentage of black or hispanic residents in a block group and the median household income of that block group.Choice of Study Areas
For the purposes of gaining insight into the phenomenon of residential segregation in American metropolitan areas, four cities were chosen. It was decided that a comparison between two cities of comparable size would be less useful than a sample of four cities of a variety of sizes, from different regions in the country. This was done with the understanding that each city has its own peculiar history, housing stock, and racial composition. It was deemed futile to choose peer cities and more appropriate to choose cities with different characteristics. With this in mind, the four cities chosen were Boston, Massachusetts, Portland, Oregon, San Diego, California, and Atlanta, Georgia.Data Acquisition and Manipulation
Both the TIGER line files and the actual 1990 Census data were obtained from Retail Profit Management's (RPM's) G.I.S. data web site, a customized interface to TIGER line boundary files and Census Data (http://home.earthlink.net/~rpminfonet/gis.html). This data needs to be customized in order to utilize it in a GIS environment.
The above data was imported in the following manner. First, the .e00 boundary files were made suitable for use as ArcView shapefiles with the Import 71 extension. These boundaries files were state-wide in their scope, but were not trimmed into smaller shapefiles until after the MSA data was brought into ArcView, so the full extent of each MSA's block group boundaries would be retained. The .csv files were downloaded in two parts for each city. The were then joined in Microsoft Excel and converted to Excel format for convenient editing and computations. Unfortunately, these data sets, while downloaded from the same source, are not easily compatible. The problem is common in ArcView; the boundary files and data files shared no common fields with which to join them. The two tables below illustrate the lack of a common field, due to the inclusion of dashes and use of different "geocode" formats.
The boundary attribute table on the left has no one "geocode" field, but rather a series of smaller numbers, which, if put together, are equivalent to the geocode field in the imported data files. Steps to make these compatible were as follows:
- Create new fields in the attribute table for the fields marked "St", "Co", "Tractbase", Tractsuf", and "Bg". Then use the field calculator to replicate the original data from the old fields in "number" rather than string form. This will make it possible perform calculations with these fields.
- Using the field calculator, construct an equation which will add these columns together to create a "unique" column. These numbers will have to be multiplied by the appropriate number of zeros, which will serve as placeholders. For instance, The new "St" column will be multiplied by 10000000000, the new county code by 10000000, etc, so that these can be added to create a 12 digit unique number.
Preliminary Analysis of Data: Visual and Quantitative
An initial analysis of the census data revealed the following numbers and percentages for the four metropolitan areas:
The lower half of the table constitutes the initial step of measuring racial isolation. Atlanta's metro area contains 280 block groups that are over 90% black, or racially homogenous. It also contains a staggering 914 block groups that are over 90% white. Boston shows similar disparity, but the numbers are different because the black population is far smaller than that of Atlanta in relation to the metro area (6% as opposed to 26%). The last two columns show these numbers as percentages of the total number of block groups in each metro area. When shown in this manner, it is easy to see that San Diego, even with 614 block groups over 90% white in population, still has the lowest degree of racial isolation. Only 38% of San Diego's block groups are racially homogenous, as opposed to over 70% in Boston. The map below is part of a preliminary examination of Boston's black enclave, south of the Downtown area in Roxbury and Dorchester. The map displays race as part of a pie chart for each block group, with the following color coding:
Red: White
Green: Asian
Blue: Black
Yellow: Hispanic
Grey: Amerind/Other
The segregation index was applied to provide a more quantitative analysis of the separation between black and white residents in the four study cities. The first step in this process was to identify all block groups that have black percentages that are higher than the metropolitan area's breakdown, and total the population counts in these areas (black and white). This is easily accomplished in ArcView using sort and calculate functions. After running the results through the formula (Duncan and Duncan's Index of Dissimilarity, see above), the following results were obtained:
The number obtained from the dissimilarity index indicates the percentage of the black population that would have to be relocated spatially (theoretically of course) to create a situation where each block group contained the same racial split at the MSA as a whole. Basically, the numbers show that the two northern cities with rather small black populations are somewhat more segregated that Atlanta and San Diego. An interesting twist occurred when San Diego's numbers were run differently. Rather than calculating the black-white segregation, the black to other segregation index was calculated. A score of 10.79 resulted, indicating that blacks in San Diego are not as segregated from other groups as they are from white residents. A possible explanation might be that particular metropolitan area's large Hispanic (~500,000, 20%) and Asian populations (~200,000, 8%) which live in proximity to blacks in greater numbers than do white residents. Further research would be required to verify the validity of this finding, for instance, more accurate segregation indices could be applied and numbers for each racial group could be calculated to find the highest and lowest levels of group to group segregation.
Boston 72.54 Portland 71.62 Atlanta 59.89 San Diego 57.75 The next step was to sort the data by percentages to get a count of the number of block groups that were homogenous by each racial category at the 50% level and the 90% level (refer to table above for these numbers). These numbers were then divided back into the total number of block groups to create an index of homogeneity. While the 90% homogeneity index is striking (again, note the similarities between Boston and Portland at ~70%), the real surprise is the percentage of block groups that are 50% homogenous. The least homogenous city by this standard is San Diego, and 95% of its block groups have one racial category above the 50% mark.
The following maps were created to display areas exhibiting black "racial isolation", these areas are 90% or more black in racial composition and often occur in close proximity to one another. Click on each for a larger map.
The maps above reflect the statistics for San Diego and Portland, which have only one such block group. It should be noted that the one block group in San Diego had only 32 residents and no income, and may be some sort of institutional setting. The above block groups contain the following population numbers:
Boston Isolation Atlanta Isolation Portland Isolation San Diego Isolation
From these numbers, it is clear that Atlanta and Boston have considerable numbers of black residents living in racial isolation, while Portland and San Diego do not exhibit this phenomenon at all.
MSA % black isolation number in black isolation Boston 19.42% 45,804 Portland <1% 102 Atlanta 38.87% 285,913 San Diego <1% 32 Correlations in SPSS
The next step was to run correlations between racial percentages and median household income in all four cities. This was done in SPSS (Statistical Package for the Social Sciences), which imports data easily from Microsoft Excel spreadsheets. An initial analysis of the data revealed that normality could not be assumed, so the data was transformed to ranked data for each category and correlated using Spearman's Rank Correlation. The image below is a screen capture from a correlation done earlier using Pearson's Product-Moment Correlation, it is provided to illustrate the SPSS interface.
Literature disagrees on the extent to which income causes segregation. The argument is that if income were largely responsible, poor whites and poor black residents would live in the same areas- they almost never do (Gillmor and Doig 1992). Other factors that perpetuate segregation include discrimination, unfair lending and selling practices in real estate, and the tendency for racial groups to cluster. By comparing correlations of race and income for each city, some variation may become apparent. The following correlations were obtained. Where results were not statistically significant, the results are omitted.
(All are significant at the .01 level)
Boston Atlanta SanDiego Portland % White with Income: .425 .510 .469 .461 % Black with Income: -.385 -.507 -.442 -.362 % Hispanic with Income: -.379 - -.495 -.282 % Asian with Income: - - - - This indicates a relationship between percentage of race within a block group and the median income of that block group. As expected after review of the literature, there is a relationship between the income of a block group and its racial composition. This merely reaffirms that white American earn more money than do minorities. It should be noted that these correlations are relatively simplistic in the realm of segregation research, which often uses complicated statistical modelling techniques. This was performed as a way to utilize SPSS, but unfortunately the research problem was not well suited for display of results in SPSS, because all of the results are MSA wide and cannot be mapped back down to the block group level.
Conclusions and Future Research
From these results the conclusions are as follows. Boston exhibits the highest level of black to white residential segregation, while San Diego is the least segregated in terms of black/white segregation, by these measures. This is probably related to the presence of substantial Hispanic and Asian populations. Atlanta had largest percentage of isolation and also the strongest negative correlation between income and percentage black population per block group. It should be noted that Atlanta also has a very large black middle class, and examinations of income levels in certain areas of Atlanta would be interesting. Even with a small black population, Portland, a small Western city, is still highly segregated. However, it doesn’t show the level of racial isolation present in larger cities. Future research could examine presence of military bases and universities, or perform more advance statistical analysis of variables that may cause segregation. Exploration of causal factors in each city’s history, economic base, and race relations could also be explored to add value to traditional quantitative findings. Another possibility would be the creation of a new index that combines various measures to get a cumulative score, with the assistance of Geographic Information Systems and a spatially modified segregation index.
Other possibilities for future research would include a different segregation index, or the use of smaller sub areas of metropolitan areas to measure segregation. Another possibility would be to use the 2000 census data to compare trends to 1990 results, and possibly to measure the presence of absence of "integrated" neighborhoods over time. Once that term was defined adequately it would be relatively easy to query the GIS to find areas in certain cities that met the definition. This project has served to familiarize the author with the extent of the complexity of factors involved in the study of segregation. This is a complicated an constantly changing phenomenon, with political and social implications. Such factors must be kept in mind when designing research projects involving segregation and a Geographic Information Systems approach.Anselin, L.(1998). "G.I.S. Research Infrastructure for Spatial Analysis of Real Estate Markets". Journal of Housing Research 9:1 113-133.
Bashi, V.I. and M.A. Hughes (1997). "Globalization and Residential Segregation by Race". The Annals of the American Academy of Political Science 551 105-121.
Berry, B.J.L. and J. Kasarda (1977). Contemporary Urban Ecology. New York: Macmillan.
Carter, W.H., M. Schill and S. Wachter(1998). "Polarisation, Public Housing and Racial Minorities in U.S. Cities". Urban Studies 35:10 1889-1911.
Duncan, O. and B. Duncan(1955). "A Methodological Analysis of Segregation Indices". American Sociological Review 20 210-217.
Ebdon, David(1985). Statistics in Geography. (2nd ed.) Oxford: Blackwell.
Edmondson, B.(1996). "Black America in 2001". American Demographics (on-line), November 1996.
Egan, K., D. Anderton, and E. Weber(1998). "Relative Spatial Concentration among Minorities: Addressing Errors in Measurement". Social Forces 76:3 1115-1121.
Ellen, I.G.(1998). "Stable Racial Integration in the Contemporary United States: An Empirical Overview". Journal of Urban Affairs 20:1 27-42.
Frey, W.H. and R. Farley(1996). "Latino, Asian, and Black Segregation in U.S. Metropolitan Areas: Are Mulitethnic Metros Different?" Demography 33:1 35-50.
Gilmor, D. and S.K. Doig(1992). "Segregation Forever?". American Demographics 14:1 48-52.
Gottdiener, Mark(1994). The New Urban Sociology. New York: McGraw-Hill.
Hartshorn, T.A.(1992). Interpreting the City: An Urban Geography. New York: John Wiley & Sons, 2nd Edition.
Immergluck, D.(1998). "Progress Confined: Increases in Black Home Buying and the Persistence of Residential Segregation". Journal of Urban Affairs 20:4 443-455.
Massey, D. and N. Denton(1988). "The Dimensions of Residential Segregation". Social Forces 67:2 281-315.
Massey, D. and N. Denton(1989). "Hypersegregation in U.S. Metropolitan Areas: Black and Hispanic Segregation Along Five Dimensions". Demography 26 373-391.
Newburger, H. "Sources of Difference in Information Used by Black and White Housing Seekers: An Exploratory Analysis". Urban Studies 32:3 445-470.
O'Hare, W.P.(1993). "America's Minorities- The Demographics of Diversity".
Sandor, G.(1994). "The 'Other' Americans". American Demographics (on-line), June 1994.
Smith, D.(1979). Geographical Perspectives on Inequality. New York: Barnes & Noble.
Smith, R.(1998). "Discovering Stable Racial Integration". Journal of Urban Affairs 20:1 1-25.
Van Kempen, R. and A. S. Ozuekren(1998). "Ethnic Segregation in Cities: New Forms and Explanations in a Dynamic World". Urban Studies 35:10 1631-1656.
Wilson, W.J.(1987). The Truly Disadvantaged: The Inner City, the Underclass and Public Policy. Chicago: University of Chicago Press.
Wong, D.W.S.(1993). "Spatial Indices of Segregation". Urban Studies 30:3 559-572.
Wong, D.W.S.(1997). "Spatial Dependency of Segregation Indices". The Canadian Geographer 41:2 128-36.
Wong, D.W.S and W.K. Chong(1998). "Using Spatial Segregation Measures in GIS and Statistical Modelling Packages". Urban Geography 19:5 477-485.
Zubrinsky, C. and L. Bobo(1996). "Prismatic Metropolis: Race and Residential Segregation in the City of the Angels". Social Science Research 25 335-374.
