Diversity and Social Segregation

Year: 1990
Geographic Unit: U.S. Counties


Several different measures of residential segregation are available form this database, built from 1990 and from 2000 US Census data. These measures are available for US counties and US Primary Care Service Areas. These measures are defined following the work of Massey and Denton (1988):

Doug Massey and Nancy Denton, “The Dimensions of Residential Segregation”, Social Forces, v 67 (2)(December 1988): pp 281-315.

Summary of domains: The authors examine 20 measures from the literature to capture each of 5 dimensions of separation. The fact that there are 5 distinct dimensions was demonstrated using factor analysis. In what follows, ‘majority’ refers to whites and ‘minority’ refers to non-whites. The five domains are:

Evenness of the spatial distribution of two groups (majority, minority) among units (tracts) in a larger area (county). We capture this with the Diversity/Theil Index. This index is sensitive to the relative size of the minority population (see below). This measure correlates highly with a simpler dissimilarity index used in the literature, but has the advantage of being able to handle more than 2 groups.

Exposure of minority groups to one another (isolation) or to majority groups (interaction), attempts to measure the experience (likelihood) of segregation. We capture this with Isolation Indices, which are interpreted as probabilities of exposure. These indices are sensitive to the relative sizes of populations, because minorities can be evenly distributed across space and still not interact with many majority members if the minority population is much larger than the majority in the area. If the minorities are a small proportion they’ll experience higher exposure regardless of the evenness of their spatial distribution. When there are only 2 groups, the sum of these indices is unity (they are probabilistic complements).

Concentration refers to the relative amount of physical space occupied by minority groups in areas. We did not construct this measure; it is expected to be highly correlated with income differences across races/ethnicities. It would reflect differences in population densities experienced by people of different races. We do not capture this dimension in our data development.

Centralization captures the degree to which a group is spatially located near the center of an urban area. We did not construct this measure; with sprawling places there is rarely a central business district anymore. We do not capture this dimension in our data development.

Clustering measures the extent to which minority and majority populations live in segregated enclaves, spatially disparate from one another. We capture this with the Index of Spatial Proximity. This index also loads with the exposure group, but correlates stronger with the clustering construct than other clustering measures.

The Isolation Index, is an exposure measure (variables named Inative, Iasian, Iblack, Ihisp, Ipacif, and Iwhite, and Ias-pac). This is a separate index for each race/ethnicity; the index is a county-level or PCSA-level measure for each race/ethnicity, summed up from tracts within the county (ZCTAs within the PCSA). It represents the propensity for a minority (non-white) member to be exposed only to other minority members. It is actually a sort of probability and ranges in value from 0 to 1. Higher values represented greater isolation/segregation. In our databases, a series of these variables have been defined to represent the perspectives of different population subgroups. These have been calculated for both 1990 and 2000 from the US Census of Populations data, at both the county and PCSA geographies. The isolation index for minority group m in the county will be equal to the proportion of the total county population that is in minority group m in the special case where each tract in the county has the same (uniform) proportion in minority group m.

Diversity Index (Theil Index) (variable named Diverse )

This diversity index is an entropy index that measures the even-ness or uneven-ness of the spatial distribution of population subgroups in tracts within counties. This can reflect the extent of racial diversity in an area. The measure varies between 0 (perfectly even: all subunits/tracts have same composition as the larger area) and 1 (perfectly uneven: all subunits/tracts have only 1 population group). In our databases this variable has been constructed for both 1990 and 2000 for all counties.

Clustering into Racial/Ethnic Enclaves (variable named Spindex)

The Index of Spatial Proximity measures the extent to which subgroup populations are clustered together in adjacent tracts, suggesting the presence of a cultural enclave. In our databases, the variables have been calculated for 1990 and 2000 for all counties. An index value of unity means that there is no differential clustering between minority and majority populations. The larger the number is, the nearer majority and minority members live together (more spatially integrated). When Spindex > 1, members of each group live closer to one another’s groups than to each other (within own groups). When Spindex <1 (unusual) members of each group live closer to each others’ group than to members of their own group.

White’s clustering measure (variable named RCL )

This is another measure of spatial clustering which compares the average distance between X minority members and Y majority members. It equals 0 when minority members display the same clustering as the majority, and is positive when minority members display greater clustering than is typical of the majority. If minority members were less clustered than the majority, then the index would be negative. Disadvantage: this index has no upper or lower bounds. The variable has been calculated for all counties, with 1990 and 2000 data.

The table below sunmmarizes these measures with names, descriptions, and years available. Only the isolation indices are available for PCSAs. All measures are available for counties, in 1990 and 2000.

variable name variable description years
Diverse Diversity Index (Theil Index): measures the even-ness or uneven-ness of the spatial distribution of population subgroups in tracts within counties (ZCTAs within PCSAs). The measure varies between 0 (perfectly even) and 1 (perfectly uneven) 1990, 2000
Ias_pac Probability that Asians/Pacific Islanders will meet other Asians/Pacific Islanders 1990
Iasian Probability that Asians will meet other Asians 2000
Iblack Probability that African Americans will meet other African Americans 1990, 2000
Ihisp Probability that Hispanics will meet other Hispanics 1990, 2000
Inative Probability that American Indians/Alaska Natives will meet other American Indians/Alaska Natives 1990, 2000
Ipacif Probability that Pacific Islanders will meet other Pacific Islanders 2000
Iwhite Probability that Whites will meet other Whites 1990, 2000
RCL White's RCL measure (see documentation) 1990, 2000
Spindex Spatial Proximity Index (see documentation) 1990, 2000

Link to Documentation


Sorry, Your browser doesn't not support the iFrame tag. You can try and view the maps at: http://sal-dev.asu.edu/geodata_map.html?shortname=diversity_1990_CNTY&ts...

Alternatively you may wish to update your browser to something a little more modern: FireFox or Safari

Download Options:
Microsoft Excel
dBase Format
ESRI ShapeFile

Analysis Options:
Map Data