Geoscience Reference
In-Depth Information
segregation indices and concentration statistics based on the tract data from each
census year.
To study how neighborhoods (census tracts) are changing, however, it is critical
to use neighborhoods that have the same boundaries over time. When boundaries
shift over time, changes in household composition may be an artifact of boundary
shifts and not of any change in where people live. For the analyses of neighborhood
change, we use data from the 2000 Decennial Census and the 2005-2009 ACS; both
use the 2000 census tract definitions. Tracts are considered part of the central city if
at least 50 % of the tract population resides within the city, or if the tract is fully
enclosed by the city.
In the 2000 and 2010 Censuses, as well as in the ACS, data on same-sex
unmarried partners may be subject to coding errors. These errors are due to Census
Bureau procedures which require recoding of Census forms that list two individuals
of the same sex as spouses, and to differences in the forms and questionnaires used to
collect Census data (O'Connell and Gooding
2007
; Black et al.
2007
; O'Connell and
Felix 2011).
3
DiBennardo and Gates (
2013
) suggest a procedure through which the
preferred (corrected) state-level estimates of same-sex partners that were released by
the Census Bureau could be adjusted to the sub-state (e.g., tract) level. We rely on
their procedure, as well as data and information contained in Gates and Steinberger
(
2009
) and O'Connell and Feliz (
2011
), to create adjusted census tract counts of gay
male and lesbian partnered households, which are used in all subsequent analyses.
We measure representation of households or other groups within a census tract
by the
proportion of the city's population of the household type or other group
residing in the census tract
.
4
By construction, these proportions sum to one for each
group in each city in each year. Also, the mean value of this proportion for the
census tracts in each city will be the same for each household group (because the
proportions for each group must sum to one for the entire city, the mean proportion
for each group within a city is one divided by the number of census tracts, which is
the same for each group in the city.) We
do not
use the more common measure of
the proportion of households in the census tract who are same-sex partner
households. We use the proportion of the city population of the group residing in
the census tract for the same reasons that these proportions are used for segregation
indices. The city proportion measure is sensitive to relative differences in spatial
concentration,
but is not sensitive to relative population sizes
. Therefore, the city
3
For most of these time periods and locations, same-sex marriage was not legally recognized. The
Census assumed in most cases of one-sex partnerships listing a spouse relationship that the spouse
identification was incorrect but the gender identification was correct. The Census changed the
coding of partnerships in these cases to same-sex partners. This approach to recoding was likely to
have incorrectly classified some two-sex couples who incorrectly identified the gender of one
partner to same-sex partnerships. Because the pool of same-sex partners accounts for less than five
percent of all partnerships, any procedure that incorrectly allocates even a very small percentage of
two-sex partnerships to same-sex ones leads to substantially greater bias in estimates for same-sex
partners than for two-sex ones.
4
See Madden (
2014
) for a study of changes in the intra-metropolitan area spatial distribution of
residents by race and income using the same measurement approach.
