Geoscience Reference
In-Depth Information
neighborhoods and there is little difference in their shifts to more ethnically diverse
neighborhoods. While there is some positive correlation in locations of gay men
and lesbians, it is small. Nationally, gay men are somewhat more likely to reside in
higher vacancy neighborhoods, but there is no evidence that is the case for lesbians.
There is no evidence that centrally located neighborhoods are more attractive to gay
male or lesbian partnered households than to all households. However, both
lesbians and gay men are less likely to shift out of densely populated neighborhoods
than other households.
Census tracts that start the decade with more gay men experience significantly
greater growth in household incomes (and, therefore, presumably housing prices)
over the next decade. Large central city census tracts that start the decade with more
gay men experience significantly greater population growth over the decade, most
notably in the Northeastern region. There is little evidence that concentrations of
lesbian households are associated with future population or income growth in the
neighborhood. Census tracts with relatively more lesbians at the start of the decade
generally see few differences in either population or income growth from those with
fewer lesbians.
There are data issues to address in future research. The biggest qualification of
these results is that they rely on only one decade's data, and a potentially anomalous
decade in which the U.S. experienced its most severe recession since the Great
Depression. Furthermore, the measure of gay presence is complicated by the
U.S. Census's erroneous recoding of unmarried partner data, so that all partnerships
reporting as married and same sex are coded as unmarried same sex partnerships.
Although the data used in these analyses were adjusted to account for this potential
recoding error, a more precisely measured group of gay partners and a less
anomalous decade may give different results.
The regression modeling approach used here may also be improved with the
implementation of a more complex spatial modeling framework. The use of a
spatial Durbin model, in which lagged values of the explanatory variables are
included as additional covariates in the model, could lead to a better-fitting
model. However, the use of such a model with census tract data is likely to involve
extremely highly correlated explanatory variables, and some of the explanatory
variables in the model (e.g., distance from downtown and the city dummies) cannot
be appropriately included in a spatial Durbin model. This type of modeling strategy
does, nevertheless, warrant further investigation.
Acknowledgement We thank Marcus Dillender, Gabrielle Fack, Gary Gates, Stephen Sheppard,
and an anonymous reviewer for helpful comments on an earlier version of this paper.
