Geography Reference
In-Depth Information
Analyzing Segr egation
To look at paterns of persistence and change in segregation ideally re-
quires data with spatial characteristics from all of the areas under study
for the entire time period. Since 1910 the United States has had a limited
set of urban small-area census data, growing to include some sixty cit-
ies in forty-nine areas by 1940. For 1990 small-area data were available
for every part of the United States. his patern of data availability has
several consequences:
1. Data are not provided for comparative areas for each decade.
2. Data are not provided for all areas for all decades.
3. Some of the variables may not be available or completely
comparable from decade to decade.
4. The boundaries from decade to decade for those areas where
small-area data exist were changed from decade to decade.
The NHGIS has developed tract boundaries for all areas of the United
States for which they existed and has also experimented with using other
methods of adding spatial data. With these advancements in data avail-
ability, this essay is able to study segregation over a larger period of time
than ever before. For each year of tracted data, the NHGIS has created
boundary files that match those reported by the Census Bureau. Be-
cause residential and population paterns shit, this means that there is
no guarantee that the tract boundaries for one year will match those in
a succeeding year. For the analysis of segregation, however, the bound-
aries current in each decade were used. For 1880 enumeration district
boundaries were used, which are the same order of magnitude in popula-
tion as are census tracts.
In this essay two partially overlapping sets of time series segregation
analyses were used. One set included all cities for which data existed
at the small-area level from 1880 through 1960 and included data on
race by tract or enumeration district (for 1880). These data used the city
boundaries that existed for each decade. Since segregation measures are
not very sensitive to changing boundaries, this does not inject much bias
into the results.
18
The second set included data on metropolitan areas for
the seven decades from 1950 through 2010 as defined in the 2000 census
