Methods & data
Gay men and gay women by US neighborhood, 2025
The map estimates how many gay men and gay women live in each US neighborhood in 2025. Here a gay man or gay woman is an adult who is attracted mostly or only to the same sex, whatever their sexual identity (Section 2). No official statistic counts people by sexual orientation for small areas. The 2020 Census does count same-sex couples living together, down to the census block. GaysWya uses those couples as the spatial signal: it splits them into male and female couples, pools them over neighboring blocks, corrects for how often people of each age live in couples at all, scales them so that the national totals match survey estimates of same-sex attraction, spreads a small part of the total according to each block’s age mix, and projects the adult population to mid-2025. The block estimates are then summed to block groups, ZIP areas, places and metro regions, and contiguous areas of high concentration are outlined as gay villages. This page documents the data, the estimator, the measures and the map construction.
1 Design principles
- Same-sex couples locate the population. The census does not ask about sexual orientation, but since 2020 it identifies same-sex spouses and unmarried partners directly (Anderson et al. 2024). Following Gates (2006) and Gates and Ost (2004), the share of the gay and lesbian population living in an area is taken to follow, in large part, the share of same-sex couples living there, once the area’s age mix is taken into account (principle 5).
- Survey prevalence sets the national totals. Couples alone undercount: most adults attracted to their own sex do not live with a same-sex partner. The national totals are therefore set to survey estimates of same-sex attraction (Section 7). For residents of households and of dorms and barracks, couples decide where 85% of each total is placed and the remaining 15% is spread according to each block’s age mix; residents of institutions get the national rate for their age (principle 5).
- Men and women are estimated separately. Male and female couples have different geographies, and attraction differs by sex, so gay men and gay women are estimated separately, each as a share of adults of their own sex.
- Neighborhood scale. Block-level census counts are small and carry deliberate noise from the Census Bureau’s disclosure avoidance (U.S. Census Bureau 2021b). Each block’s couples are therefore pooled with those of all blocks within ⅓ mile, and no geography finer than the block group is shown.
- Partnering differs by age. Few same-sex couples can mean few gay adults, or few people living in couples of any kind, as around universities, where students are young and many live in dorms. Couples are therefore compared with the number expected from each neighborhood’s age mix and household population. Residents of institutions such as prisons and nursing homes, who cannot be census couples and are not part of the surrounding community, are given the national rate for their age.
- Everything adds up. Every area’s figures are sums of the same block estimates, so men and women add to the total at every level, and smaller areas add up to larger ones where they nest.
2 Terms
- Gay men are adult men (18 and over) who say they are attracted mostly or only to men. Gay women are adult women who say they are attracted mostly or only to women. Adults attracted equally to both sexes are not counted.
- The definition is by attraction, one of the three dimensions of sexual orientation alongside identity and behavior (Federal Interagency Working Group on Improving Measurement of Sexual Orientation and Gender Identity in Federal Surveys 2016), and does not depend on how people describe themselves. It therefore includes some people who identify as bisexual or straight, and leaves out some who identify as gay or lesbian. It does not use “men who have sex with men” or “women who have sex with women”, which describe behavior (Young and Meyer 2005).
- Gay adults (on the map: “Gay share”, “Gays per sq mi”, “Gay adults”) means gay men plus gay women.
- On the map, the Who picker chooses all gay adults, gay men only, or gay women only.
- Same-sex couple means a household whose householder lives with a same-sex spouse or same-sex unmarried partner. Census counts are of households, so a couple living in someone else’s household is not counted.
3 Data sources
The map covers the 50 states and the District of Columbia. Puerto Rico is left out because the 2025 projection inputs cover only the states.
| Input | Source |
|---|---|
| Householder’s spouse or partner by type (opposite- or same-sex, married or unmarried); population in institutional and non-institutional group quarters, 8.13 million blocks | 2020 Census DHC table P17, “Household Type (Including Living Alone) by Relationship” (U.S. Census Bureau 2023b), via IPUMS NHGIS (Manson et al. 2023) |
| Adult men and women by age | DHC table P12 (Schroeder et al. 2025) |
| Total population, population in households by age (under 18, 18 and over), land area, ZIP Code Tabulation Area | DHC tables P1 and P15 (Manson et al. 2024) |
| Male-male and female-female couple households (married and unmarried), 84,414 tracts | DHC table PCT15, “Coupled Households, by Type” (Schroeder et al. 2025) |
| County population, April 2020 base and July 2025 | Vintage 2025 county estimates, CO-EST2025-POP (U.S. Census Bureau, Population Division 2026a) |
| Housing units by block, 2020 and July 2025 | Address Count Listing Files (U.S. Census Bureau 2022, 2025) |
| Block shapes (for distances between blocks) | 2020 TIGER/Line block geodatabase (U.S. Census Bureau 2021a) |
| Places (“allij”: place-based block-group clusters), regions (“bslm”: regions similar to BEA economic areas), simplified block-group shapes | Crosswalks and boundaries from the author’s synth-us-county project (New York City split into its five boroughs) |
| ZIP area shapes | 2020 ZIP Code Tabulation Areas, 1:500,000 cartographic boundary file (U.S. Census Bureau 2023a) |
| Sexual attraction by sex and age group, adults 18 and over (294,690 respondents) | National Survey on Drug Use and Health public use files, 2015–2019 and 2022–2023 (Center for Behavioral Health Statistics and Quality 2020, 2024) |
| National adults by sex and age, July 2025 | NC-EST2025-AGESEX-RES (U.S. Census Bureau, Population Division 2026b) |
| Adults living in a same-sex couple, by sex and age group, household adults (6.8 million person records) | American Community Survey 2022 and 2023, via IPUMS USA (Ruggles et al. 2025) |
4 Notation
Blocks are indexed by \(b\), and \(T(b)\) and \(c(b)\) are the tract and county containing block \(b\). The superscript \(s \in \{m, f\}\) marks quantities for men and women.
- \(S_b\): same-sex couple households in block \(b\), the count of persons who are the householder’s same-sex spouse or same-sex unmarried partner (P17). Each couple household has exactly one such person.
- \(M_b\), \(F_b\): adult men and women in 2020, the P12 totals for each sex minus the four age groups under 18.
- \(\pi_T\): the male share of same-sex couple households in tract \(T\) (PCT15).
- \(\gamma_b\): the block’s 2020–2025 growth factor (Section 6), and \(M'_b = \operatorname{round}(\gamma_b M_b)\), \(F'_b = \operatorname{round}(\gamma_b F_b)\) the projected 2025 adults.
- \(n^s_{b,a}\): adults of sex \(s\) in age group \(a\) (18–25, 26–34, 35–49, 50–64, 65 and over) in block \(b\) in 2020, from P12. P12’s 25–29 group is split one-fifth to 18–25 and four-fifths to 26–34.
- \(H_b\): adults living in households (P15), and \(h_b = \min(1, H_b / (M_b + F_b))\) the household share of the block’s adults.
- \(\iota_b\): the share of the block’s adults in institutional group quarters, \((1 - h_b)\) times the institutional share of its group-quarters residents (P17).
- \(N(b)\): the neighborhood of block \(b\), all blocks within ⅓ mile of it, including itself (Section 8).
- \(G^m_b\), \(G^f_b\): estimated gay men and gay women in block \(b\) (Section 8).
5 Couples by sex
Block tables do not give the sex of same-sex couples, but the tract table PCT15 does. Each block inherits its tract’s male share of same-sex couples,
\[ \pi_T = \frac{\text{MM}^{\text{married}}_T + \text{MM}^{\text{unmarried}}_T}{\text{MM}^{\text{married}}_T + \text{MM}^{\text{unmarried}}_T + \text{FF}^{\text{married}}_T + \text{FF}^{\text{unmarried}}_T}, \]
\[ \text{MM}_b = \pi_{T(b)}\, S_b, \qquad \text{FF}_b = \bigl(1 - \pi_{T(b)}\bigr) S_b . \]
The tract shares are used as published, without smoothing. In the 1,903 tracts with no same-sex couples, \(\pi_T\) is undefined and the national value is used instead, \(\pi_{\text{nat}} = \sum_b \pi_{T(b)} S_b / \sum_b S_b = 0.476\). Nationally, slightly fewer same-sex couples are male than female.
The 2020 Census was the first to offer separate answer categories for same-sex and opposite-sex spouses and partners, together with an automated check of sex against relationship on electronic forms (Anderson et al. 2024). Earlier censuses inferred same-sex couples from the reported sexes, so a small error rate among the far more numerous opposite-sex couples inflated the count: an estimated 28% of same-sex couple households in the 2010 Census were opposite-sex couples (O’Connell and Feliz 2011).
6 Adults and the 2025 projection
Rates are computed on 2025 adults. The projection uses the change in each block’s housing units, scaled so that every county matches the Census Bureau’s 2025 population estimate.
County growth. For each county, \(g_c = P^{\text{Jul 2025}}_c / P^{\text{Apr 2020 base}}_c\) from CO-EST2025-POP. Counties are matched by name, ignoring accents, spacing and punctuation. Connecticut’s eight 2020 counties, which Vintage 2025 replaces with planning regions, take the state’s growth factor.
Block housing growth. For each block with housing units in 2020, \(v_b = \operatorname{clip}\bigl(U^{\text{Jul 2025}}_b / U^{2020}_b,\; 0.1,\; 10\bigr)\), where \(U\) is the total housing units in the Address Count Listing Files. Block housing-unit counts are among the few 2020 census figures published without disclosure-avoidance noise (U.S. Census Bureau 2023b).
County anchoring. Within each county, the housing ratios are rescaled so that their 2020-population-weighted mean equals the county’s growth:
\[ \gamma_b = \operatorname{clip}\!\left(v_b\, \frac{g_{c}\, \sum_{b' \in c} P_{b'}}{\sum_{b' \in c} v_{b'} P_{b'}},\; 0.1,\; 10\right)\!, \qquad c = c(b), \]
where \(P\) is 2020 population and the sums run over the county’s blocks that have a housing ratio. Blocks with no housing units in 2020 (28% of blocks, mostly unpopulated or with group quarters only, together holding 0.8% of adults) take \(\gamma_b = g_{c(b)}\). Without the final clip, the population-weighted mean of \(\gamma_b\) in every county is exactly \(g_c\); the clip moves a few sparsely settled counties slightly.
Projected adults and population. \(M'_b = \operatorname{round}(\gamma_b M_b)\) and \(F'_b = \operatorname{round}(\gamma_b F_b)\). The same factor is applied to both sexes, so each block’s sex and age composition is held at 2020. “Adults” on the map is \(M' + F'\). “Population” is the 2020 total population of the area scaled by the growth of its adults. For that ratio, adults are counted as household adults (P15) scaled up to the total population, so the ratio measures growth only.
7 Calibration to national totals
7.1 Survey data
The only US probability survey that asks adults of all ages about sexual attraction and releases the answers in public files is the National Survey on Drug Use and Health (Center for Behavioral Health Statistics and Quality 2020, 2024). It interviews about 56,000 people a year, drawn from the civilian, noninstitutionalized population. Since 2015 it has asked: “People are different in their sexual attraction to other people. Which statement best describes your feelings?” The answers run from only attracted to the opposite sex, through mostly opposite, equally, and mostly same, to only same; “I am not sure” is also offered. The most recent public file with the question is 2023.
- Answers counted. “Mostly same” and “only same”. “Not sure” and missing answers (3.9% of adults) are left out of the denominator.
- Two periods. In 2015–2019 the survey was in person, with sensitive questions answered privately on the interviewer’s computer. From late 2020 it moved to a mix of web and in-person interviews, and reported same-sex attraction jumped. 2020 and 2021, the transition years, are skipped. Within each period the years are pooled with their survey weights.
- Averaging. For each sex and age group, the rates of the two periods are averaged with equal weight. The national target is the average of these age-group rates, weighted by the July 2025 adult population of each sex (U.S. Census Bureau, Population Division 2026b).
| 2015–2019 | 2022–2023 | Target (2025 ages) | |
|---|---|---|---|
| Gay men: men attracted mostly or only to men | 2.77% | 3.87% | 3.32% |
| Gay women: women attracted mostly or only to women | 2.34% | 3.20% | 2.77% |
The design-based standard errors of the targets are 0.10 and 0.07 percentage points. Rates fall with age:
| Age | Men | Women |
|---|---|---|
| 18–25 | 4.4% | 5.1% |
| 26–34 | 4.4% | 3.8% |
| 35–49 | 3.2% | 2.7% |
| 50–64 | 3.6% | 2.3% |
| 65 and over | 1.7% | 1.5% |
Other surveys agree in pattern. The National Survey of Family Growth, which covers only younger adults, found about 2.3% of men and 1.6% of women aged 18–44 attracted mostly or only to the same sex in 2011–2013 (Copen, Chandra, and Febo-Vazquez 2016), and about 2.7% and 2.9% at ages 18–49 in 2015–2019 (National Center for Health Statistics, n.d.). An earlier face-to-face survey with a show card, the 2004–2005 National Epidemiologic Survey on Alcohol and Related Conditions, found 1.9% and 1.7% (McCabe et al. 2009). Adults attracted equally to both sexes, who are not counted here, are a further 1.5% of men and 5.0% of women in the same NSDUH data. Reported rates rise over time and are lower in interviewer-administered surveys, and self-reports of sexual orientation are generally understated (Coffman, Coffman, and Ericson 2017). The targets average the in-person period and the web period rather than choosing between them.
7.2 Two parts of each target
Each target is split into two parts that are placed differently:
| Allocated by couples, \(\tau^s\) (85%) | Spread by age, \(\alpha^s\) (15%) | Total | |
|---|---|---|---|
| Gay men (share of adult men) | 2.82% | 0.50% | 3.32% |
| Gay women (share of adult women) | 2.35% | 0.42% | 2.77% |
- Most of the target follows same-sex couples. Gay men and lesbians often live with a same-sex partner: 37–46% of gay men and 51–62% of lesbians aged 18–59 in California data (Carpenter and Gates 2008), and about a third of gay and lesbian adults nationally (Jones 2025). In a national probability survey, LGBT people who did not live with a partner lived in places similar to those who did (Lee et al. 2020), so the couples are a reasonable guide to where the unpartnered live too.
- A small part is spread by age. It gives every neighborhood a floor, for adults the couples do not locate: people without a same-sex partner who live apart from couples, people not open about their orientation, and places where couples are under-reported. Same-sex attraction falls with age (table above), so this part follows each block’s age mix (\(e^s_b\) below). It changes only the floor: for men it ranges from about 0.3% of adults in a block of people aged 65 and over to about 0.7% in a block of young adults.
- The 85/15 split is a modeling choice. No data measure what share of the population couples fail to locate. It is not the share living with a partner, which is much lower; in 2020 only 0.89% of adult men lived with a male spouse or partner, about a quarter of the men’s target.
7.3 Age-corrected couples signal
The number of same-sex coupled people in an area depends on how many gay adults live there, but also on how often anyone there lives in a couple. Students rarely do, and people living in group quarters (dorms, prisons, nursing homes) never count as census couples. Around universities, few same-sex couples per adult therefore do not mean few gay adults. The signal compares each neighborhood’s same-sex coupled people with the number expected from its age mix.
Expected couples. Let \(c^s_a\) be the national share of household adults of sex \(s\) and age group \(a\) who live in a same-sex couple, as householder, spouse or unmarried partner. It is estimated from the 2022 and 2023 American Community Survey (Ruggles et al. 2025):
| Age | Men | Women |
|---|---|---|
| 18–25 | 0.44% | 0.87% |
| 26–34 | 1.33% | 1.76% |
| 35–49 | 1.21% | 1.26% |
| 50–64 | 1.15% | 0.94% |
| 65 and over | 0.62% | 0.52% |
Each block’s observed same-sex coupled people, and those expected from its household adults, are
\[ O^m_b = 2\,\text{MM}_b, \quad O^f_b = 2\,\text{FF}_b, \qquad X^s_b = h_b \sum_a c^s_a\, n^s_{b,a}, \]
with \(h_b\) removing adults who live in group quarters, who cannot be census couples. With \(q^s_a\) the national attraction rate for sex \(s\) and age group \(a\) (Section 7, table above), the attraction rate expected from the block’s age mix is
\[ e^s_b = \frac{\sum_a q^s_a\, n^s_{b,a}}{\sum_a n^s_{b,a}} . \]
Signal. Pooling over the neighborhood \(N(b)\),
\[ x^s_b = \frac{\sum_{b' \in N(b)} O^s_{b'}}{\sum_{b' \in N(b)} X^s_{b'}} \times \frac{\sum_{b' \in N(b)} e^s_{b'}\, n^s_{b'}}{\sum_{b' \in N(b)} n^s_{b'}}, \qquad n^s_{b'} = \sum_a n^s_{b',a} , \]
is the ratio of observed to expected same-sex coupled people, times the attraction rate expected from the neighborhood’s age mix. In a neighborhood with the national age mix and no group quarters, \(x^s_b\) is proportional to same-sex coupled people per adult. Where few couples are expected, each couple counts for more.
Group quarters. The signal describes the neighborhood’s own residents. It applies to each block’s local adults: those in households and in non-institutional group quarters (college dorms, military barracks, group homes, shelters), so students in dorms share their neighborhood’s rate. Dorm and barracks residents count as local only where at least 400 adults live in households within ⅓ mile; with fewer, a single couple nearby would set the rate for a much larger group-quarters population, so they get the national rate for their age instead. Adults in institutional group quarters (prisons, jails, nursing homes, juvenile and psychiatric facilities) are not part of the surrounding community; they get the national attraction rate for the block’s age mix, \(e^s_b\) (above). So do local adults in neighborhoods with no households, where no couples are expected and the signal is undefined. Group-quarters residents are assumed to share their block’s age mix. The local share of a block’s adults is \(\ell_b = 1 - \iota_b\) (or \(h_b\) where fewer than 400 household adults live within ⅓ mile), and \(\ell_b = 0\) where the signal is undefined.
7.4 Logit shift
The signal is first rescaled so that its national mean over local adults equals the national number of same-sex coupled people per household adult, so that the shift compresses the highest values as it would for that plain rate. The couples part of the target for local adults, \(\tau^s_\ell\), is then reached with a single shift \(\delta^s\) on the logit scale, chosen so that the mean over local adults hits it:
\[ \frac{\sum_b w^s_b\, \sigma\!\left(\operatorname{logit} x^s_b + \delta^s\right)}{\sum_b w^s_b} = \tau^s_\ell, \qquad \sigma(x) = \frac{1}{1 + e^{-x}}, \quad w^m_b = \ell_b M'_b,\; w^f_b = \ell_b F'_b . \]
\(\tau^s_\ell\) is set so that, together with the institutional and other national-rate adults, the national totals equal the targets (\(\tau^s + \alpha^s\) over all adults); because few adults get national rates, it differs only slightly from \(\tau^s\). Before the shift, values are bounded to \([10^{-8},\, 1 - 10^{-8}]\), and \(\delta^s\) is found by root-finding. A shift on the logit scale keeps every rate between 0 and 1 and preserves the order of blocks. It is the one-constraint case of calibration with a logit distance function (Deville and Särndal 1992). For small values the shift amounts to multiplying by the same factor, \(e^{\delta^s}\), so neighborhoods with no same-sex couples stay near zero and the target is allocated roughly in proportion to the signal.
7.5 Age-weighted part
The age part is added to every block according to its own age mix:
\[ \alpha^s_b = \alpha^s\, \frac{e^s_b}{\bar e^s}, \qquad \bar e^s = \frac{\sum_b w^s_b\, e^s_b}{\sum_b w^s_b}, \]
with \(\bar e^s\) the mean over local adults, so that their national mean of \(\alpha^s_b\) is exactly \(\alpha^s\). Blocks without adults take \(\alpha^s\). The calibrated rate over all of a block’s adults is
\[ \hat r^s_b = \ell_b \min\!\left(\sigma\!\left(\operatorname{logit} x^s_b + \delta^s\right) + \alpha^s_b,\; 1\right) + (1 - \ell_b)\, e^s_b . \]
8 Neighborhoods and block estimates
Couples are pooled over each block’s neighborhood \(N(b) = \{\, b' : d(b, b') < \tfrac13 \text{ mile} \,\}\), which includes the block itself:
- Distance \(d\) is the haversine (great-circle) distance between interior points of the 2020 block polygons. Neighbors can lie across tract, county and state lines. Candidate pairs were drawn from blocks in the same or nearby tracts (tract internal points less than 2 miles apart), so in large rural tracts a few neighbors across tract lines can be missed.
- Pooling adds up counts before dividing, so a block with one or two couples does not produce an extreme ratio.
- Blocks with no neighbors in the table (3,211 blocks, mostly unpopulated) use their own counts.
Pooling evens out the small counts and disclosure-avoidance noise of individual blocks; the Census Bureau advises aggregating block data before use (U.S. Census Bureau 2021b). It also reflects that a neighborhood’s character extends beyond a single block. Because calibration comes after pooling, the national rates equal the targets: 3.32% for men and 2.77% for women.
Block estimates. \(G^m_b = \hat r^m_b M'_b\) and \(G^f_b = \hat r^f_b F'_b\). They are kept as fractions, not rounded per block, so that sums over many small blocks are not biased. Figures are rounded only for display. Nationally, the estimates total 4.28 million gay men and 3.79 million gay women in 2025, out of 128.9 million adult men and 136.9 million adult women.
9 Measures
For any area \(a\) (block group, ZIP area, place, metro region or village), let \(G^s_a = \sum_{b \in a} G^s_b\), \(M'_a\) and \(F'_a\) be the summed adults, and \(L_a\) the land area in square miles.
| Measure | All | Gay men | Gay women |
|---|---|---|---|
| Gay share | \((G^m_a + G^f_a) / (M'_a + F'_a)\) | \(G^m_a / M'_a\) | \(G^f_a / F'_a\) |
| Gays per sq mi | \((G^m_a + G^f_a) / L_a\) | \(G^m_a / L_a\) | \(G^f_a / L_a\) |
| Gay adults | \(G^m_a + G^f_a\) | \(G^m_a\) | \(G^f_a\) |
The combined share is the adult-weighted mean of the two sex-specific shares. Popups show all three versions of each measure with the counts of adults and population.
9.1 Gay Prevalence Index
The Gay Prevalence Index combines concentration and density into a score from 0 to 100. It is computed for every block group in the country and for each of the three versions (all, gay men, gay women):
\[ \text{GPI}_u = 100\left(\tfrac23\, R_w(\text{share}_u) + \tfrac13\, R_w(\text{density}_u)\right), \]
\[ R_w(x_u) = \frac{\sum_{u'} w_{u'}\,\mathbb{1}\{x_{u'} \text{ ranked below } x_u\} + \tfrac12 w_u}{\sum_{u'} w_{u'}}, \]
where share and density are the gay share and gays per square mile of block group \(u\), and the weight \(w_u\) is its adults of the relevant sex (both sexes for “All”). \(R_w\) is the weighted mid-rank: the share of the nation’s adults living in block groups that rank lower, plus half of the block group’s own. Ties are rare and are ordered arbitrarily. A GPI of 90 thus means that the block group’s two ranks, weighted ⅔ and ⅓, average to the 90th percentile of adults. Block groups with no adults of the relevant sex have no index.
For larger areas, the index is the adult-weighted mean of the block-group indexes: places and metro regions average their block groups, and ZIP areas, which do not nest block groups, average over their blocks, each block carrying its block group’s index. It describes the typical adult’s surroundings on the same scale at every level.
10 Areas
| Areas | Description | Count drawn |
|---|---|---|
| Metro regions | Regions (“bslm”); titled “Name Urban Region” when they span several counties | 2,768 |
| Places | Places (“allij”); New York City is split into its five boroughs | 14,520 |
| ZIP areas | 2020 ZIP Code Tabulation Areas | 33,296 |
| Block groups | 2020 block groups | 237,975 |
- Places carry their name and state; unincorporated areas read “Unincorporated County, ST”.
- ZIP areas are the Census Bureau’s block-based approximations of ZIP codes (U.S. Census Bureau 2023a). Each block is assigned to the ZCTA recorded for it in the census tables; every block in the 50 states and DC has one. A ZIP area’s popup names the place holding most of its adults.
- Areas with no coupled households of any kind (opposite- or same-sex) are not drawn.
11 Gay villages
Gay villages, or “gayborhoods” (Ghaziani 2014), are contiguous areas where gay men, gay women or both are concentrated. A block group is a candidate if any of six conditions holds:
- the gay share is at least 10%; or
- the share of gay men, or of gay women, is at least 15%; or
- the combined Gay Prevalence Index is at least 97.5; or
- the index for gay men, or for gay women, is at least 99.
The sex-specific conditions are stricter than the combined ones, so they add only areas where one group is concentrated on its own. A block group is also a candidate if the place containing it meets any of the six conditions as a whole.
Candidate block groups that share a boundary segment (rook adjacency; a shared corner is not enough) are joined into connected areas. Holes inside an area are filled for display, but the village’s figures cover only its candidate block groups. Areas with fewer than 500 gay adults are dropped. Each remaining village is named after the place holding most of its gay adults, numbered when a place has more than one, with a few names set by hand (for example, “Greater West Hollywood-Silver Lake, CA”). The map shows 139 villages, home to 650,000 gay adults.
12 Map construction
- Tiles. All levels and the villages are PMTiles vector tiles served with the site.
- Gay adults. Under this measure each area is a circle, drawn at a precomputed interior point, whose area is proportional to its count. The scale differs by level, so circles are comparable within a level, not across levels.
- Clicking an area opens its popup and outlines it in black. With Gay Villages switched on, clicking a village opens the village’s popup instead.
- Tables. The same figures for every metro region, place, ZIP area and village are listed on the Tables page.
13 Limitations
- Couples are a proxy. Partnered gay men and lesbians are disproportionately white and highly educated (Carpenter and Gates 2008), so places where such couples live may be overstated relative to places where unpartnered gay men and gay women live.
- National, not local, prevalence. The targets fix only the national totals. The model assumes that the relation between couples and the wider population is the same everywhere, and that the part not placed by couples varies across places only through age.
- Self-reported attraction. Answers depend on how the question is asked; the jump between interview modes shows how much. The survey covers the civilian, noninstitutionalized population, so its rates are applied to residents of institutions and military barracks without being measured for them.
- Group quarters. Institutional residents get national rates, and group-quarters residents are assumed to share their block’s age mix; neither is measured locally.
- National partnering rates by age. The age correction assumes that, at a given age, people partner as often everywhere as nationally. Where people of the same age partner more or less often for other reasons, the signal is too high or too low.
- Sex split at tract level. Within a tract, every block has the same male share of same-sex couples.
- Mixed dates. Couples are counted in April 2020; adults are projected to July 2025 with each block’s 2020 sex and age composition; partnering rates are from 2022–2023.
- Disclosure-avoidance noise. Block counts include deliberate noise (Abowd et al. 2022; U.S. Census Bureau 2021b). Pooling and aggregation reduce, but do not remove, its effect on the smallest areas.