1 Introduction

Since 2000, Census data have allowed researchers to track the segregation of same-sex couple households from other households. Spring (2013) demonstrated that the segregation of both male and female same-sex couples from different-sex couples declined within the 100 largest cities in the United States between 2000 and 2010. Some scholars have taken this as evidence that same-sex couples are dispersing out of established “gayborhoods.” However, multiple demographic processes could be driving declines in segregation. For several reasons, same-sex couples might be making up lower shares of all households in gay neighborhoods, while retaining relatively static representations in other parts of the city. Or, same-sex couples might be increasingly represented in new parts of the city, as they grow in numbers and/or move from other areas. Thus, the fact alone that segregation has declined in many cities leaves open many unanswered questions about how exactly this pattern emerged.

Also important are the characteristics of changing neighborhoods. What features define neighborhoods that are losing shares of same-sex partners? Urbanicity? High housing costs? And what features define neighborhoods that are gaining shares of same-sex partners? Suburban locations? Affordable housing? Racial and ethnic diversity?

To answer these questions, this chapter takes a closer look at metropolitan statistical areas (MSAs) with declining segregation of male same-sex households and describes the sociodemographic characteristics of neighborhoods within these MSAs. Using data from the 2000 and 2010 US Censuses, this chapter identifies neighborhoods that had increasing, decreasing, or static shares of male same-sex households. Demographic, geographic, and socioeconomic characteristics of these neighborhoods are then compared. This chapter also investigates changes in gay neighborhoods, including changes in their prevalence and spatial distribution within MSAs. Demographic, geographic, and socioeconomic characteristics of neighborhoods that transitioned to or from a gay neighborhood, or remained unchanged, are also compared. Lastly, this chapter investigates the small number of MSAs that experienced increasing segregation of male same-sex households and explores neighborhood changes within these areas as well.

Through such descriptions, a better picture of within-MSA changes in the spatial demography of male same-sex households can emerge, providing clues into the future geographies of same-sex partners. At the same time, such descriptions call for us to be critical consumers of census data, as these data are not very inclusive and notoriously error-prone for same-sex partners (DiBennardo and Gates 2014). The chapter concludes with a look ahead to the 2020 Census, and what it will mean for future research on the geography and segregation of same-sex partners.

2 Data and Methods

2.1 Decennial Census Data

Data come from the 2000 and 2010 US Decennial Census Summary File 1, the 100% census count of the US population (US Census Bureau 2000, 2010). Census tracts boundaries are used to approximate neighborhoods. Data are standardized to geographic boundaries from the year 2000 (Geolytics Inc. 2010) to account for any shifts in census tract boundaries over the study time period. Male same-sex partners are identified in the census data by combining individual responses to two questions: the sex of each individual living in the household and their relationship to the household head. Two males who report an “unmarried partner” relationship are defined as male same-sex partners. “Unmarried partner” is defined by the census as a person who was not related to the householder but who had a “close personal relationship” with them. The Census Bureau recoded responses of “same-sex spouse” to “unmarried partner” over the study time period.

There are several limitations to this census-based definition of same-sex partners. First, the Census did not ask sexual orientation directly, so it can only be inferred for individuals living with unmarried partners of the same sex. This means the census data do not represent the entire LGBTQ population. Second, some same-sex partners may have been unwilling to identify themselves on the Census. Census follow-up studies have assessed the undercount of same-sex partners and have estimated that 16–19% of same-sex partners did not identify themselves in Census 2000 (Badgett and Rogers 2003), and 10% of same-sex partners did not identify themselves in 2010 (Gates 2010). Third, there were substantial errors in the 2010 Census that resulted in a significant number of different-sex partners being counted as same-sex partners (O’Connell and Feliz 2011). The Census Bureau re-estimated the number of same-sex partners using micro-data level files of respondents’ first names and an index of the sex commonly associated with their first names (see O’Connell and Feliz 2011 for a detailed description of the methodology). Revised counts for 2000 and 2010 were released by the Census Bureau at the state level. Using the procedure outlined by Gates and Cooke (2012), I then apply the state error rates to individual census tracts to calculate revised tract estimates. The revised census tract estimates are used throughout this analysis including in the calculation of segregation scores. Fourth, the present analysis relies on census tracts to represent neighborhoods, even though the boundaries of census tracts are somewhat arbitrary. Different definitions of neighborhoods could lead to segregation estimates that are higher or lower, especially if these definitions differ dramatically from census tract boundaries.

2.2 Segregation Scores

This study categorizes MSAs as those that experienced declining or increasing segregation of male same-sex partners from different-sex partners (including married and unmarried partners) between 2000 and 2010. I measure segregation with the index of dissimilarity. The index of dissimilarity compares two groups at a time, and values represent the percentage (ranging from 0 to 100) of one group needing to change residences (in this case, move into a census tract where they are underrepresented) in order to achieve an even distribution (Duncan and Duncan 1955).

The index of dissimilarity is statistically independent of the relative size of the two groups used in its computation, which is particularly important in this study because different-sex partners greatly outnumber same-sex partners in all MSAs. However, the index can be sensitive if the population of one group is small compared to the number of census tracts used in its calculation (Johnson and Farley 1985). To ensure there is a substantial number of male same-sex partners for analysis, this study is limited to the 100 most populous MSAs.Footnote 1 Segregation indices cannot be reliably calculated for smaller MSAs.

2.3 Gay Neighborhoods

This study relies on a demographic definition of gay neighborhoods. To determine whether a census tract is a gay neighborhood, the tract percent of male same-sex households out of all households is compared to a threshold for the metropolitan area. The threshold adopted for this analysis is at or above the 90th percentile for percent male same-sex households in the MSA in the year 2000.Footnote 2 Tracts are compared to the 2000 threshold in 2000 and 2010, to determine whether tracts were a gay neighborhood in 2000 and whether tracts transitioned into or out of gay neighborhood status by 2010. This threshold is somewhat arbitrary and could just as easily be set at a lower or higher percentile. The 90th percentile is used because it generates a large enough number of gay neighborhoods to calculate summary statistics, without being overly inclusive. This purely demographic definition does not encompass the full identity and broader symbolic meaning of gay neighborhoods; for that I direct readers to other excellent chapters in this volume.

2.4 Other Neighborhood Characteristics

Data on other census tract characteristics are drawn from the 2000 and 2010 US Decennial Census (US Census Bureau 2000, 2010) and the 2005–2009 American Community Survey (US Census Bureau 2009) and are also standardized to year 2000 geographies. Of particular interest to this study is the urbanicity of neighborhoods. I define “urban” neighborhoods as those that are inside the principal city(ies) of their metropolitan area.Footnote 3 Neighborhoods outside of a principal city are defined as “suburban.” Other tract variables include median home values, share of housing units that are owner-occupied, total population, share of households with own children under 18, share of the population age 25+ with a college degree, median income, and racial-ethnic distributions. All monetary variables are specified in year 2010 dollars, and any comparisons made to the year 2000 are adjusted for inflation.

2.5 Analysis

I first summarize segregation scores in 2000 and 2010, dividing the sample of MSAs into those that experienced declining segregation of male same-sex households and those that experienced increasing segregation. I then take a descriptive look at the neighborhoods within each of those MSA contexts. I describe the share of tracts within each group of MSAs that experienced declining, increasing, or static shares of male same-sex partners out of all tract households, and the share of neighborhoods that transitioned to or from a gay neighborhood. Then, within each of those neighborhood categories, I summarize the sociodemographic characteristics of those neighborhoods.

3 Results

3.1 Trends in Segregation and Neighborhood Change

The dominant pattern in US metropolitan areas is declining segregation of male same-sex households. Among the 100 largest MSAs in the United States, 96 experienced declining levels of segregation between 2000 and 2010 and only 4 experienced increasing levels of segregation. Table 2.1 shows the average index of dissimilarity across these MSAs. In desegregating MSAs, the index of dissimilarity was 61.73 in 2000 and 52.10 in 2010, a decline of 9.63 points. In increasingly segregated MSAs, the index of dissimilarity was 51.00 in 2000 and 54.09 in 2010, an increase of 3.09 points. What is notable about these numbers is that desegregating MSAs started out at higher levels of segregation than did MSAs where segregation increased. In fact, both groups of MSAs end up at similar levels of segregation, with segregation scores that are in the low 50s. According to guidelines from Massey and Denton (1993), the overall average index of 52.18 is considered in the “moderate” range for segregation.

Table 2.1 Segregation of male–male households from male–female households, 2000–2010

The levels of segregation reported in Table 2.1 are higher than segregation scores reported in Spring (2013), primarily because Spring (2013) based that assessment on individual cities (i.e., census-designated places), while the data here are for MSAs. This suggests that the inclusion of the surrounding suburban areas of MSAs contributes to higher segregation scores for male–male households. The differing geographies of male same-sex households in urban versus suburban areas is an important point I return to throughout the chapter.

Because declining segregation is far and away the dominant trend and only four MSAs actually experienced increasing segregation, I now turn to more in-depth assessment of desegregating MSAs. Within the context of a desegregating MSA, how have individual neighborhoods changed?

Table 2.2 reports changes in neighborhood shares of male same-sex households out of all households. Within the context of desegregation, tracts on average increased their shares of male–male households, going from an average of 0.18% male–male households in 2000 to 0.26% male–male households in 2010. Although these percentages are still quite small (they are both less than one percent), the trend that is driving declining segregation appears to be increasing representation of male same-sex households across most neighborhoods. Indeed, Table 2.2 also shows that more than half of all tracts (57.79%) within the 96 desegregating MSAs had increasing shares of male–male households. By comparison, 27.93% of tracts experienced declining shares of male–male households, and 14.27% experienced no change.

Table 2.2 Change in tract shares of male–male households within desegregating MSAs

Another way to assess neighborhood-level change in desegregating MSAs is to compare demographic changes in gay neighborhoods. Table 2.3 shows how many neighborhoods met the demographic threshold for a gay neighborhood in 2000 and 2010. By definition, about 10% (9.89%) of neighborhoods were defined as gay neighborhoods in 2000 (since the definition is based on meeting the 90th percentile for tract share of male same-sex households in the MSA in 2000). By 2010, 16.51% of neighborhoods were defined as gay neighborhoods, suggesting that an expanding number of gay neighborhoods is another important contributor to declining segregation.

Table 2.3 Transitions of gay neighborhoods within desegregating MSAs*

Despite this expansion, most neighborhoods (80.30%) were not gay neighborhoods in 2000 or 2010. However, a fairly substantial number of neighborhoods (9.8%) transitioned to gay neighborhood status by 2010. Of the remaining neighborhoods, 3.19% transitioned out of being a gay neighborhood, while 6.71% were gay neighborhoods in both time periods. Overall the data in Table 2.3 somewhat challenge the idea that gay neighborhoods are declining or withering away. A small minority of gay neighborhoods did transition away from having substantial shares of male same-sex households. But more common was that gay neighborhoods remained gay neighborhoods, and that new gay neighborhoods emerged.

3.2 Characteristics of Neighborhoods Within Desegregating Cities

The previous section demonstrated the divergent pathways of neighborhoods within desegregating cities. While many neighborhoods have expanded their shares of male same-sex households, a few have gone in the other direction, and many others have remained unchanged. What are the sociodemographic characteristics of neighborhoods following each of these trends?

Table 2.4 focuses on urbanicity, housing, and other demographic characteristics of tracts within desegregating MSAs, depending on whether the tracts experienced declining, increasing, or static shares of male same-sex households. Data are reported for 2010. The level and direction of change from 2000 is also reported.

Table 2.4 Characteristics of tracts within desegregating MSAs, by direction of change in the share of male–male households

Among tracts with declining shares of male same-sex households, the average share of male–male households out of all households was 0.16% in 2010, representing a 0.16 percentage point decline from the year 2000. This means that in neighborhoods that lost shares of male same-sex partners, shares of male–male partners were on average cut in half. About half (50.55%) of these neighborhoods were located in urban areas and another half were located in suburban areas (49.45%). Since about 45% of tracts in the sample of MSAs were in urban areas, neighborhoods that lost shares of male–male households were somewhat overrepresented in urban areas. These neighborhoods also represent areas where housing values have grown (+$40,842), median incomes have grown (+$931), and homeownership rates were lower than in other neighborhoods (54.30%).

Neighborhoods with increasing shares of male same-sex households seem to have absorbed about the same percentage of male–male households that were lost from declining neighborhoods. These neighborhoods nearly doubled their shares of male same-sex households from 2000 to 2010, and by 2010 had the greatest shares of male same-sex households compared to other neighborhoods. These neighborhoods were about equally distributed between urban and suburban areas. They too had increasing home values (+$40,121) and were otherwise fairly similar to neighborhoods with declining shares of male same-sex partners. These neighborhoods may have offered a few more opportunities for homeownership and had slightly higher rates of college degrees and higher median incomes. These neighborhoods also appeared to be less diverse than neighborhoods with declining shares of male same-sex partners. The former were 61.32% white, while the latter were 53.30% white.

Finally, Table 2.4 reveals that neighborhoods with static shares of male same-sex partners were static because they had no male–male households in either time period. These tracts were overwhelmingly suburban (77.35%). These tracts also had greater increases in home values (+$65,004) than other neighborhoods, and much higher rates of homeownership (72.37%), more households with children (26.64%), higher median incomes ($71,156), and were more white (74.94%). These data suggest that what appear to be quintessentially suburban neighborhoods have remained closed-off to male same-sex partners, even within a larger MSA context of declining segregation.

The second way this chapter categorizes neighborhoods is by the transition of gay neighborhood status. Table2.5 shows summary statistics for neighborhoods depending on whether, across the two time points, the neighborhoods were never a gay neighborhood, remained a gay neighborhood, or transitioned one way or the other. Those that were gay neighborhoods in 2000 and remained gay neighborhoods in 2010 had the highest shares of male same-sex households (1.35%) compared to all other neighborhoods. These neighborhoods were very likely to be urban neighborhoods (88.32% urban) and were substantially underrepresented in the suburbs (11.68% suburban). These neighborhoods also had the highest and most rapidly increasing home values ($345,420; up $101,483 since 2000), along with the lowest rates of homeownership (42.12%), compared to all other neighborhoods. It is important to note that despite high housing costs, these neighborhoods remained gay neighborhoods and even increased their shares of male same-sex households. These neighborhoods also had low rates of households with children (13.44%), high rates of college degrees (44.85%), some of the highest incomes ($57,860), and were not especially diverse (61.01% white, non-Latinx).

Table 2.5 Characteristics of tracts within desegregating MSAs, by transitions of gay neighborhoods*

The least common neighborhood was one that transitioned from a gay neighborhood in 2000 to a non-gay neighborhood in 2010. In these neighborhoods, the average tract share of male same-sex households was 0.20% in 2010, down 0.45 percentage points from 2000. These were mostly urban neighborhoods (70.54%), but also some suburban (29.46%). These neighborhoods, on average, had lower housing values than other neighborhoods ($244,482), the lowest rates of college degrees (25.13%), and the lowest median incomes ($46,028), suggesting that these neighborhoods were perhaps struggling socioeconomically. These neighborhoods were also the most racially diverse neighborhoods, with higher than average shares of Black and Latinx residents (27.09 and 21.13%, respectively).

Some neighborhoods became gay neighborhoods over the time period. These neighborhoods saw the highest increases in shares of male same-sex households (+0.46 percentage points), compared to other neighborhoods. These neighborhoods were 67.98% urban and 32.02% suburban. When it comes to housing values and other demographic characteristics, these neighborhoods tended to be neither the highest nor lowest compared to the other neighborhoods. These neighborhoods were moderate in terms of home values, rates of homeownership, incomes, and other sociodemographic characteristics. This seems to suggest that new gay neighborhoods did represent a departure from those that were already gay neighborhoods in 2000, in that new gay neighborhoods were more “average” types of neighborhoods (average, that is, demographically).

Not surprisingly, those that were never gay neighborhoods had the lowest shares of male same-sex households (0.13%), compared to all other neighborhoods. These neighborhoods were more often in the suburbs (62.34%), but a substantial share were also urban (37.66%). These neighborhoods had the highest rates of homeownership (61.34%), the highest rates of households with children (23.19%), the highest median incomes ($60,600), and the lowest levels of racial diversity (62.15% white), compared to all other neighborhoods. Much like the tracts with zero shares of male same-sex partners described in Table 2.4, these typically well-off, mostly suburban neighborhoods have remained relatively closed-off to male same-sex households.

3.3 Increasingly Segregated Cities

As previously shown, the dominant trend among MSAs is declining segregation of male same-sex partners. However, among the 100 most populous MSAs, 4 MSAs experienced increasing segregation. These were Los-Angeles-Long Beach-Santa Ana, CA; Stockton, CA; Miami-Fort Lauderdale-Pompano Beach, FL; and Augusta-Richmond County, GA-SC. This section takes a brief look at neighborhood changes within these increasingly segregated metropolitan areas.

Although only 4 MSAs experienced increasing segregation, these MSAs included over 4,000 census tracts. Table 2.6 assesses whether the distribution of male same-sex households across these 4,000+ tracts differed from the average distribution in desegregating MSAs. The data indicate that tracts in increasingly segregated MSAs had, on average, greater shares of male–male households than in desegregating MSAs in both 2000 and 2010. More of the neighborhoods in increasingly segregated MSAs had declining shares of male–male households than in desegregating MSAs (5.65% more), while fewer neighborhoods had increasing shares (1.51% fewer). These differences are fairly modest, but may indicate that in increasingly segregated MSAs there was a trend of male same-sex households leaving a large number of neighborhoods to settle (or resettle) in a smaller number of neighborhoods. In other words, while the trend in desegregating MSAs was toward neighborhood “deconcentration” of male same-sex partners, the trend in increasingly segregated MSAs seems to be more toward (re)concentration.

Table 2.6 Neighborhood change within increasingly segregated MSAs*

The change in gay neighborhoods in increasingly segregated MSAs follows along the same lines. In 2000, desegregating and increasingly segregated MSAs had similar numbers of gay neighborhoods. But by 2010, increasingly segregated MSAs had fewer gay neighborhoods than desegregating MSAs. This is primarily because more neighborhoods in increasingly segregated cities (2.64% more) remained ‘not gay’ in 2000 and 2010.

Because this is a summary of only four metropolitan areas, and the MSAs themselves are very different in terms of geography and socioeconomic conditions, readers are cautioned against reading too much into the data in Table 2.6. Further statistical summary of these four MSAs likely would not carry much meaning (which is why I do not present that here), but further research on increasingly segregated cities could follow several important directions. First, it might be interesting to assess what these cities have in common with cities that only slightly declined in segregation. The line that separates these types of cities might be very small, and perhaps these cities as a group are quite distinct from cities that are more rapidly desegregating. Such comparisons could be useful for identifying the metropolitan-level conditions that support more rapid, as opposed to more minimal, changes in segregation, since the dominant trend is toward decline anyway. Second, this statistical work should be complemented by qualitative descriptions of cities with varying contexts of segregation. The four MSAs identified as increasingly segregated might provide interesting and important counterpoints to qualitative research in desegregating cities.

4 The Demographic Future of Gay Neighborhoods

What do the aforementioned results mean for the demographic future of gay neighborhoods? Contrary to some popular assumptions, gay neighborhoods are not ceasing to exist. However, gay neighborhoods are demographically changing and spatially reorganizing, even within the broader context of declining segregation. Underlying declining rates of segregation seems to be the increasing suburbanization of male same-sex households and gay neighborhoods.

At the same time, some suburban areas (and some urban areas, to a lesser extent) have remained closed-off to male same-sex households. These tend to be economically vibrant, “child-friendly,” mostly suburban neighborhoods. It will be interesting to see if these neighborhoods remain unchanged in the 2020 census. It is probably unlikely that these neighborhoods will remain isolated from the dominant trend of increasing male same-sex populations at the neighborhood-level. However, quantitative and qualitative researchers may find it interesting to track how these neighborhoods change, and how same-sex households fare when they enter these neighborhoods.

5 Future Research: Census 2020 and Beyond

Future research on the segregation of same-sex partners depends on the availability of high-quality data. The census is a federal product mandated by the constitution, and is arguably the most important and consequential source of data on the US population. One could argue that the accuracy with which the census counts LGBTQ populations says a lot about society. If the United States wants to strive for equality based on gender identity and sexual orientation, it must also strive for accuracy in national data collection efforts. Accurate estimates matter for both practical reasons (e.g., the allocation of funding for community resources) and symbolic reasons (e.g., the visibility of marginalized communities), which is why there is so much at stake for getting the estimates right.

Due to inaccuracies with Census 2010, Census 2020 moved to a new way to count same-sex partners. In 2000 and 2010, same-sex partnerships were inferred by cross-referencing the sex of each person and their relationship to the household head. This method was prone to error, because as it was later revealed, a significant number of individuals mis-marked their sex and were erroneously counted as same-sex partners (O’Connell and Feliz 2011). The new method asked directly within the relationship question whether each person was an “Opposite-sex husband/wife/spouse,” an “Opposite-sex unmarried partner,” a “Same-sex husband/wife/spouse,” or a “Same-sex unmarried partner” (US Census Bureau 2019b). This reframing of the question should considerably improve the accuracy of same-sex partner population counts and is a huge step in the right direction.

What Census 2020 still lacked was a separate LGBTQ category. Detailed questions that ask directly about sexual orientation and gender identity would finally allow single LGBTQ people to be counted. The US Census Bureau reportedly proposed such questions in the lead-up to the 2020 Census (Wang 2018). However, those questions were quickly removed from consideration by the federal administration shortly after Trump took office in 2017. In 2018, the “Census Equality Act” was introduced in the US Senate, which, if it becomes law, would require sexual orientation and gender identity questions to be added to the census by 2030 and the American Community Survey by 2020 (Govtrack 2019). According to the bill’s sponsor Senator Kamala Harris, “the spirit of the census is that no one should go uncounted and no one should be invisible” (Govtrack 2019).

To obtain accurate data, the Census Bureau not only needs to ask the right questions, it also need to overcome peoples’ concerns about participation. Misinformation about the census, concerns over confidentiality of responses, and general distrust in the government all serve as barriers to participation. A number of advocacy organizations are seeking to break through those barriers, encouraging LGBTQ populations to “get out the count” and “queer the census” (National LGBTQ Task Force 2019). Yet even with 100% participation, the census can still only provide a simplistic, point-in-time snapshot of LGBTQ populations. LGBTQ identities can be complex, in flux, and individualized (Browne 2010); and as such, there will always be some misrepresentation when the only option is to check a box.

Despite its limitations, the US Census remains a critical source of data on same-sex households. Other useful sources of data include large scale social surveys like the General Social SurveyFootnote 4 and the National Longitudinal Study of Adolescent to Adult Health Footnote 5—which have continued to fold more LGBTQ people into their samples, and smaller targeted studies like the Williams Institute’s mixed-methods Pathways to Justice studyFootnote 6—which focus specifically on LGBTQ people. Such expansion of qualitative and quantitative data on LGBTQ people will be instrumental in furthering our understanding of LGBTQ lives. But the Census remains the only data source large enough in scale to track the geographic segregation of same-sex households over time. For scholars interested in how the spatial reorganization of LGBTQ populations impacts LGBTQ communities, such demographic estimates of segregation and change are important for setting the broader context. The need for such research further underscores the need for a non-political, accurate, and inclusive national census.