Geoscience Reference
In-Depth Information
gay households, we also report parallel regression analyses for the census tracts'
2009 shares of all households in the central city.
16
The results do not support all of the hypotheses asserted in the literature on the
location of gay households in cities.
Diversity
Gates and Ost's contention that gays are attracted to more diverse
neighborhoods is not confirmed by the patterns of shifts in gay locations between
2000 and 2009 in cities in any region of the country, at least to the extent that racial
composition is used as a measure of diversity. Census tracts with greater shares of the
central city's African American population in 2000 saw greater decreases in their
share of the city's gay male partnered population by 2009 than in their share of all
households, nationally and within each region. A similar result was found for lesbians
in the Northeast and Midwest. Notably, lesbians in Western cities disproportionately
shifted toward census tracts with larger African American populations.
If we measure diversity using Hispanic composition of neighborhoods, gay male
partnership households are less likely than all households to shift away from census
tracts with more Hispanic households. Lesbian partnership households, however,
are slightly more likely to shift away from high Hispanic concentration census
tracts in the Midwest and Northeast.
Vacancies and Older Housing
Collins (
2004
) and Ruting (
2008
) contend that gays
are attracted to neighborhoods with high vacancy rates and older housing stock.
Although their discussion does not exclude lesbians explicitly, their examples apply
to gay men. Table
19.3
provides some support for the hypothesis that gay men shift
toward census tracts with higher vacancy rates, and of being disproportionately in
census tracts with older housing in Western, Midwestern and Northeastern cities,
relative to all households. While the results do not contradict the hypothesis that
gay men are attracted to high vacancy neighborhoods, the differences with all
households are not very big. Lesbians are also more attracted than all households
to neighborhoods with older housing. Lesbian partnered households are also less
likely to be attracted to high vacancy census tracts; they are significantly less likely
16
In general, the coefficients from a spatial autoregressive model cannot be directly compared to
the coefficients from a spatial error regression, as the lagged dependent variable in the
autoregressive model introduces feedback and indirect impacts. While the coefficient for an
explanatory variable (X) in a spatial error model is interpreted as the average impact of X on
the outcome (Y), the average impact of X on Y in the spatial autoregressive model is calculated
based on the coefficient, the spatial weight matrix, and the spatial autoregressive term [
in
Eq. (
19.2
)] (LeSage and Pace
2009
). In the results presented here, the feedback and indirect
impacts in the autoregressive models are quite small, and do not make a substantive difference in
the conclusions reached. As such, only the coefficients themselves are presented in the results
tables.
ˁ
