Association between same-sex partnership certificate schemes and mental health in sexual and gender minority adults: nationwide cross-sectional study in Japan

Abstract

Objectives

Same-sex marriage legalisation improves mental health among sexual and gender minority (SGM) people. In Japan, where same-sex marriage is not legalised, some municipalities have introduced ‘partnership certificate’ schemes (PCSs) to offer same-sex couples social recognition without legal recognition. We examined whether mental health status in SGM adults differed between municipalities with and without PCSs.

Design, setting and participants

We used cross-sectional data on adults aged ≥18 years from a nationwide internet survey conducted in September and October 2022. We estimated the relative difference by comparing the differences in mental distress and suicidal thoughts reported by SGM adults residing in municipalities with and without PCSs to the corresponding differences among cisgender heterosexual adults.

Primary and secondary outcome measures

Mental distress (Kessler 6 Scale score ≥5) and self-reported suicidal thoughts, adjusted for sociodemographic characteristics, health status and municipal characteristics.

Results

We analysed 28 106 participants (mean (SD) age, 48.2 (17.3) years), including 3169 (11.3%) SGM individuals. Despite no significant difference in mental distress, we found a statistically significant difference in rates of self-reported suicidal thoughts between municipalities with and without the scheme for the SGM group (average marginal effects (AME), +5.1 percentage points (pp); 95% CI +0.4 to +9.8 pp) and the cisgender heterosexual group (AME, –1.6 pp; 95% CI –3.0 to –0.1 pp; relative difference, +6.7 pp; 95% CI +1.8 to +11.5 pp).

Conclusions

Residing in a municipality with PCSs was not associated with better mental health among SGM people in Japan. This may reflect limited scheme uptake or perceptions of inequality with legal marriage. These schemes without national legislation may not adequately address internalised stigma or promote acceptance, and they may have lowered the self-esteem of SGM people, leading to worse mental health. Further research is needed to clarify the underlying mechanisms of this association.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• The study used nationwide data from a large-scale survey of adults aged ≥18 years, allowing for a comprehensive assessment that reflects the Japanese population.

• To minimise the effects of unmeasured confounding factors, we calculated a relative difference by comparing the difference in mental health status between sex and gender minority adults in municipalities with and without partnership certificate schemes with the corresponding difference in the cisgender heterosexual group.

• The use of cross-sectional data limited causal inference and prevented the assessment of temporal ordering between the partnership certificate scheme and mental health outcomes.

• The cross-sectional data limited our ability to capture delayed mental health effects, suggesting the need for a longitudinal follow-up study.

Introduction

Sexual and gender minority (SGM) people experience significant health disparities, particularly concerning mental health. Research consistently shows that SGM individuals have a higher risk of mental disorders than their cisgender heterosexual counterparts.¹,³ Social stressors, such as discrimination from family or peers and stigma due to negative societal attitudes,^{4 5} contribute to higher rates of depression, anxiety, substance use and suicidality among SGM people.⁶,⁸

More than 30 countries have legalised same-sex marriage, mostly in Europe and the Americas,⁹ since the Netherlands became the first in 2001.¹⁰ Legalisation provides same-sex couples with social and welfare benefits once exclusive to different-sex married couples, such as tax exemptions.¹¹ Additionally, studies show a marked decline in both implicit and explicit homophobic biases following legislation,¹² suggesting that same-sex marriage legislation affects public opinion and the social recognition of SGM people by eliminating this institutional discrimination.¹³ Studies also report reduced perceived stigma and discrimination,¹⁴ as well as improved mental health among SGM people, after the legalisation.¹⁵ The legal recognition of same-sex relationships mitigates minority stress and improves the mental health of SGM people by enhancing societal validation and reducing structural inequities.¹⁶

In Asia, same-sex marriage is only legal in Taiwan, Nepal and Thailand.^{9 17} Although Japan has not yet legalised same-sex marriage as of 2026, growing advocacy for the welfare of SGM people and the elimination of discrimination has led many local governments to introduce non-legally binding ‘partnership certificate’ schemes (PCSs) since 2015.^{18 19} These schemes differ from legal different-sex marriage, especially regarding inheritance, parental authority and tax deductions. However, the schemes attempt to offer social recognition and visibility to these relationships by providing an official acknowledgement, helping same-sex couples to more easily explain their relationships in situations such as housing applications, hospital visits and interactions with employers, in addition to offering access to local government-level welfare as though they were married, including for public housing and accident compensation payment. Since its introduction in Tokyo in 2015, the scheme expanded to approximately one quarter of Japan’s 1741 municipalities by September 2022 (online supplemental figures S1 and S2). Although the response to this scheme among the SGM population was largely positive,²⁰ use has remained limited, with around 3500 couples nationwide having received certificates by September 2022 (online supplemental table S1).^{21 22}

These schemes are considered the next-best alternative where legalising same-sex marriage is politically difficult. Whereas the legalisation of same-sex marriage has been associated with mental health improvement by reducing institutional discrimination and stigma,¹⁴,¹⁶ PCSs typically provide limited legal rights and leave persistent structural inequalities unaddressed, which may be insufficient to alleviate minority stress. Furthermore, although same-sex marriage legalisation in the USA did not trigger significant backlash,²³,²⁵ local initiatives without national legalisation may provoke public opposition against SGM through increased visibility, potentially diminishing their mental health benefits. Therefore, we hypothesised that PCSs might not be associated with better mental health among SGM individuals.

In Japan, assessing the impact of PCSs on SGM mental health is important for informing the schemes’ expansion and future legalisation debates. Simple comparisons of SGM mental health outcomes between municipalities with and without PCSs may be influenced by other unmeasured municipal-level differences. Therefore, we adopted a study design that compared these differences among SGM individuals with the corresponding differences observed among cisgender heterosexual individuals, under the plausible assumption that such municipal-level factors similarly affect mental health in both groups. We thus aimed to examine whether residing in municipalities with PCSs was associated with mental health outcomes in SGM populations.

Methods

Data

We analysed cross-sectional data from the Japan COVID-19 and Society Internet Survey (JACSIS) 2022, a web-based, self-reported questionnaire survey conducted by a major nationwide internet research agency with 2.3 million panellists.²⁶ These voluntary panellists are members of Rakuten, Japan’s largest e-commerce company, which provides a wide range of online and offline services, including travel, mobile carrier services and financial services. The survey was sampled via stratification based on sex, age and prefecture category to reflect national demographics. Questionnaires were distributed to 76 363 panellists aged 16–82 years from 12 September 2022 to 19 October 2022, and closed on reaching 32 000 participants (the response rate was 41.9%). Quotas were set by age, sex (registered in the agency’s database) and prefecture category.²⁷

Population

We excluded 3370 individuals with irregular or inconsistent responses to three predefined questions, as in other studies using JACSIS data (online supplemental methods).^{27 28} Additionally, 511 individuals with missing residential zip codes and 13 individuals aged <18 years were excluded, leaving 28 106 respondents (87.8%) eligible for analysis.

Study design

In a cross-sectional comparison of the mental health status of SGM adults in municipalities with and without PCSs, resident characteristics and other municipality-level policies might differ in unmeasurable ways and could affect mental health, leading to confounding bias. To minimise unmeasured confounding bias, the difference in mental health status between SGM adults in municipalities with and without PCSs was compared with the corresponding difference in the cisgender heterosexual group (relative difference or difference-in-differences).

Given that the introduction of PCSs was unlikely to be associated with the mental health of the cisgender heterosexual population, any observed differences would indicate the magnitude of confounding bias due to unobserved differences in resident characteristics or municipality-level policies.

SGM status

SGM status was defined by the answers to two questions in the survey: (1) ‘How do you describe your own gender (gender identity)?’ and (2) ‘Which gender are you romantically or sexually attracted to (sexual orientation)?’ Each question offered five options: ‘men,’ ‘women,’ ‘both men and women,’ ‘unable or unwilling to decide,’ and ‘others.’

Participants self-identifying as ‘women’ and attracted to ‘women’ were defined as lesbian, those self-identifying as ‘men’ and attracted to ‘men’ were defined as gay, and those attracted to ‘men and women’ were defined as bisexual. Participants self-identifying as ‘men’ and attracted to ‘women’ and those self-identifying as ‘women’ and attracted to ‘men’ were defined as heterosexual. Those whose gender identity was ‘both men and women’ and those whose gender identity or sexual orientation was ‘unable or unwilling to decide’ or ‘others’ were defined as ‘uncategorised.’ These differ from common definitions in Western countries,^{29 30} but the questionnaires were created considering differences in sociocultural backgrounds.³¹

The participant’s sex at birth was determined using the same five response options. Participants whose sex at birth and gender identity were inconsistent (‘men’ to ‘women’ or vice versa) were categorised as transgender; those with consistent identities (‘men’ or ‘women’) were classed as cisgender. Individuals selecting ‘unable or unwilling to decide’ or ‘others’ for either attribute were labelled ‘uncategorised.’ Although some people may be both transgender and either lesbian, gay or bisexual, we defined SGM if any one of lesbian, gay, bisexual, transgender or uncategorised was met, and all others were defined as cisgender heterosexual (online supplemental table S2).

Exposure variable

In line with studies in the USA and the Netherlands,^{15 32} this cross-sectional study aimed to assess the association between the introduction of a PCS at the municipal level and the mental health of all SGM individuals—regardless of whether they had a partner—rather than the association among those issued a certificate. Using zip codes, we identified whether each respondent’s municipality had introduced a PCS by September 2022.³³

Outcome variables

The outcome variables were mental distress and self-reported suicidal thoughts. Mental distress was defined by a Kessler 6 Scale score of ≥5.³⁴,³⁶ Suicidal thought was defined as responding ‘yes’ to the following question: ‘Have you ever felt that you wanted to die in the last year?’ Participants could choose one of five mutually exclusive responses. The five possible responses were (1) ‘Yes, for the first time in the last 2 months,’ (2) ‘Yes, for the first time in the last year,’ (3) ‘Yes, not for the first time but in the last year,’ (4) ‘No, but experienced before 1 year ago,’ and (5) ‘Never.’ Participants who answered with options 1–3 were defined as having suicidal thoughts.

Adjustment variables

We adjusted for the characteristics of the participants and the municipalities in which they lived. Participants’ characteristics comprised age, sex at birth, marital status, academic attainment, employment status, household income level, presence of comorbidity, walking disability and self-reported health (details are described in online supplemental methods).

Municipal characteristics were derived from the 2020 Population Census and statistics on local taxes from the Ministry of Internal Affairs and Communications.^{37 38} They comprised the following continuous variables: population; taxable income per taxpayer (Japanese yen); population density; and proportions of children (aged <15 years), older people (aged ≥65 years), women and foreign residents. These variables were standardised and included up to a cubic term in the model to allow for nonlinear relationships.

Statistical analysis

First, the characteristics of SGM and cisgender heterosexual adults were described. To address sampling and non-response bias, an inverse probability weighting approach was applied throughout the analyses.³⁹ Although the study population’s distributions of age group, sex and prefecture categories were similar to national ones, it may not be nationally representative due to potential selection bias in the internet-based survey. Therefore, the inverse of propensity scores representing the estimated probability of responding to the survey was calculated using the 2019 Comprehensive Survey of Living Conditions.^{40 41}

Second, mental health outcomes were compared separately between the SGM and cisgender heterosexual groups and between municipalities with and without PCSs (without using any interaction terms between SGM status and PCSs). A multivariable logistic regression model adjusted for participants’ and municipalities’ characteristics was used. Standard errors were clustered at the municipality level to account for a potential correlation of participants living in the same municipality.⁴² The adjusted mental distress rate was calculated using marginal standardisation.⁴³ To improve the interpretability of findings, the average marginal effects (AMEs) of each SGM status and the PCS introduction were calculated (instead of ORs), representing the differences in rates of mental distress and suicidal thoughts.

Third, the associations between the introduction of PCSs and SGM adults’ mental health outcomes were assessed using the same regression models with the interaction term between SGM status and PCS introduction in residential municipalities. The following were calculated and reported: (1) the AME of the PCS introduction, separately for the SGM and cisgender heterosexual groups; and (2) the relative difference (difference-in-differences; differences in the AME between the SGM and cisgender heterosexual groups). These difference-in-differences were calculated by comparing the differences in mental health among SGM people with and without PCSs to similar differences in the cisgender heterosexual group. This was to minimise the effect of unmeasured confounding factors because people living in municipalities with and without PCSs may differ in unmeasured characteristics affecting mental health.

Fourth, to examine whether the association between living in municipalities with PCSs and mental health outcomes varied by sexual orientation, multivariable logistic regression analyses were performed with the interaction between sexual orientation (lesbian, gay, bisexual, uncategorised and heterosexual) and PCS.

Finally, because the mental health of the SGM population might not be affected immediately after the introduction of a PCS, the analysis was repeated by categorising the participants according to the time since the introduction (not introduced, introduced in the past 2 years, or introduced more than 2 years ago). As a sensitivity analysis, we repeated the analysis using a 4-year cut-off instead of a 2-year cut-off. Two-tailed p values of <0.05 were interpreted as statistically significant. Stata V.17 (Stata Corp.) was used.

Secondary analyses

We conducted several subgroup analyses. First, analyses were repeated stratifying by age (<45 vs ≥45 years), marital status (married vs unmarried) and parenthood (living with their children vs not living with or not having children), given the potentially greater psychological impact of PCSs on younger, unmarried or childless SGM individuals.⁴⁴ Second, analyses were repeated stratifying by self-reported history of mental illness (depression or other psychological disorder).

We also conducted several sensitivity analyses. First, since PCSs were expected to have a particularly significant impact on SGM individuals with same-sex partners, we repeated analyses by restricting the SGM group to gay or lesbian individuals with a same-sex partner. Second, given the substantial variation in residential municipalities’ PCS utilisation proportions (certificates issued per million residents), we conducted a sensitivity analysis in which participants living in municipalities with PCS were subdivided into two groups: high utilisation (≥50 per million residents) and low utilisation (<50 per million residents). Third, we divided the ‘uncategorised’ SGM into two groups (unable or unwilling to decide and others), considering the possibility of misclassification. We performed multivariable logistic regression analyses with the interaction between sexual orientation (lesbian, gay, bisexual, unable or unwilling to decide, others and heterosexual) and PCSs.

Patient and public involvement

No respondents were involved in setting the research question or the outcome measures, nor were they involved in the design, implementation or interpretation of the study. Informing the study participants of the results was not possible because we used de-identified data. The results will be disseminated through this publication and other media.

Results

Characteristics of the study population

We analysed 28 106 adults (mean (SD) age, 48.2 (17.3) years), comprising 3169 (11.3%) SGM adults and 24 937 (88.7%) cisgender heterosexual adults (table 1). Among them, 14 647 (52.1%) resided in 415 municipalities with PCSs, and 13 459 (47.9%) resided in 961 municipalities without PCSs. The municipalities with PCSs tended to have a large population, high population density, low proportion of older people and high taxable income (online supplemental table S3). SGM adults were younger, less likely to be married and more likely to have comorbidities, and had lower household incomes than cisgender heterosexual adults. Based on the partner’s sex reported by participants, approximately half of married SGM individuals were in different-sex marriages (online supplemental table S4), and most lived with their spouses (online supplemental table S5).

Table 1. Participant characteristics by the presence of a partnership certificate scheme in the residential municipality.

Living in municipalities with partnership certificate schemes	Sexual and gender minority (N=3169)				Cisgender heterosexual (N=24 937)
	Yes (N=1612)	(%)	No (N=1557)	(%)	Yes (N=13 035)	(%)	No (N=11 902)	(%)
Age, mean (SD), years	41.6	(17.1)	43.3	(17.4)	48.8	(17.3)	49.0	(17.0)
Sex at birth
Women	774	(48.0)	765	(49.1)	6415	(49.2)	6350	(53.3)
Men	632	(39.2)	596	(38.3)	6574	(50.4)	5506	(46.3)
Not answered/others	206	(12.7)	196	(12.6)	45	(0.3)	46	(0.4)
Sexual orientation
Heterosexual	67	(4.2)	63	(4.1)	13 035	(100)	11 902	(100)
Lesbian	292	(18.1)	330	(21.2)	0	(0)	0	(0)
Gay	416	(25.8)	382	(24.5)	0	(0)	0	(0)
Bisexual	239	(14.8)	216	(13.9)	0	(0)	0	(0)
Uncategorised	598	(37.1)	565	(36.3)	0	(0)	0	(0)
Transgender status
Cisgender	902	(56.0)	884	(56.7)	13 035	(100)	11 902	(100)
Transgender	112	(6.9)	108	(7.0)	0	(0)	0	(0)
Uncategorised	598	(37.1)	565	(36.3)	0	(0)	0	(0)
Marital status
Married	610	(37.8)	648	(41.6)	8462	(64.9)	7785	(65.4)
Never married	818	(50.7)	794	(51.0)	3295	(25.3)	2798	(23.5)
Widowed	113	(7.0)	56	(3.6)	576	(4.4)	645	(5.4)
Divorced	72	(4.4)	58	(3.8)	701	(5.4)	675	(5.7)
Academic background
Junior high school/high school	831	(53.3)	908	(60.8)	6866	(53.3)	6623	(56.2)
College*	285	(18.3)	248	(16.6)	2465	(19.1)	2380	(20.2)
University/graduate school	443	(28.4)	337	(22.6)	3553	(27.6)	2776	(23.6)
Household income level†
Low	542	(33.6)	542	(34.8)	3953	(30.3)	3721	(31.3)
Medium	278	(17.2)	271	(17.4)	3564	(27.3)	3204	(26.9)
High	252	(15.7)	207	(13.3)	2705	(20.8)	2378	(20.0)
Not answered	540	(33.5)	536	(34.4)	2812	(21.6)	2599	(21.8)
Employment status
Employer	116	(7.2)	76	(4.9)	366	(2.8)	392	(3.3)
Self-employed	99	(6.1)	83	(5.4)	832	(6.4)	642	(5.4)
Employee	957	(59.4)	882	(56.6)	7365	(56.5)	6924	(58.2)
Unemployed	441	(27.3)	516	(33.2)	4472	(34.3)	3944	(33.1)
Comorbidities
Hypertension	297	(18.4)	247	(15.9)	2219	(17.0)	1936	(16.3)
Diabetes	155	(9.6)	116	(7.5)	865	(6.6)	772	(6.5)
Asthma	111	(6.9)	76	(4.9)	342	(2.6)	258	(2.2)
Myocardial infarction	109	(6.8)	137	(8.8)	267	(2.0)	227	(1.9)
Stroke	84	(5.2)	80	(5.1)	119	(0.9)	104	(0.9)
Chronic obstructive pulmonary disease	80	(5.0)	77	(4.9)	92	(0.7)	80	(0.7)
Cancer	83	(5.2)	61	(3.9)	266	(2.0)	246	(2.1)

The numbers are no. (%), except for age. We used inverse probability weighting for the probability of participating in our survey; therefore, the numbers of high-trust and low-trust individuals do not necessarily add up to the total number of participants.

^*College includes junior, community and vocational colleges.

^†Income level indicates the tertiles of household equivalent income (‘low’=less than 2.5 million Japanese yen (JPY), ‘medium’=2.5–4.4 million JPY and ‘high’=more than 4.4 million JPY).

Mental health status among SGM and cisgender heterosexual individuals

We observed that 50.4% of the SGM group (1597/3169) and 35.4% of the cisgender heterosexual group (8831/24 937) had mental health distress; 797 SGM (28.4%) and 3223 cisgender heterosexual (11.5%) individuals had suicidal thoughts. After adjusting for participant and municipality characteristics, SGM adults were significantly more likely to express mental distress (AME, +5.5 percentage points (pp); 95% CI +2.4 to +8.6 pp) and suicidal thoughts (AME, +4.0 pp; 95% CI +1.7 to +6.4 pp) than cisgender heterosexual adults (table 2). Living in municipalities with PCSs was not significantly associated with mental distress (AME, 0.6 pp; 95% CI –1.4 to +2.5 pp) or suicidal thoughts (AME, –0.8 pp; 95% CI –2.2 to +0.6 pp).

Table 2. Association between sexual and gender minority status or the introduction of a partnership certificate scheme and mental health conditions.

	No. of participants	Adjusted rates* % (95% CI)	Average marginal effect* pp (95% CI)
Mental distress
Sexual orientation
Sexual and gender minority	3169	42.2 (39.2 to 45.1)	5.5 (2.4 to 8.6)
Cisgender heterosexual	24 937	36.6 (35.7 to 37.6)	Reference
Municipality of residence
With partnership certificate scheme	14 647	37.5 (36.1 to 38.9)	0.6 (−1.4 to 2.5)
Without partnership certificate scheme	13 459	36.9 (35.7 to 38.1)	Reference
Suicidal thoughts
Sexual orientation
Sexual and gender minority	3169	17.8 (15.7 to 19.9)	4.0 (1.7 to 6.4)
Cisgender heterosexual	24 937	13.8 (13.1 to 14.5)	Reference
Municipality of residence
With partnership certificate scheme	14 647	13.9 (13.0 to 14.9)	−0.8 (−2.2 to 0.6)
Without partnership certificate scheme	13 459	14.8 (13.9 to 15.6)	Reference

We adjusted for age, sex at birth, marital status, academic attainment, employment status, household income level, comorbidities, walking disability, self-reported health and several municipal characteristics. SEs were clustered at the municipality level. The average marginal effects of mental distress and self-reported suicidal thoughts are shown for each analysis. Adjusted rates were calculated using marginal standardisation.

^*To examine the association between sexual and gender minority status or introduction of a partnership certification scheme and each mental health outcome (mental distress (Kessler 6 Scale score ≥5) or self-reported suicidal thoughts), we used separate multivariable logistic regression models.

pp, percentage point.

PCS and mental health outcomes of SGM adults

We found no evidence that living in municipalities with PCSs was associated with mental distress in SGM adults (relative difference, +4.6 pp; 95% CI –1.2 to +10.3 pp; figure 1 and online supplemental table S6). No difference was shown in the proportion of mental distress between municipalities with and without PCSs among either the SGM group (AME, +4.6 pp; 95% CI –0.9 to +10.2 pp) or the cisgender heterosexual group (AME, +0.07 pp; 95% CI –2.0 to +2.1 pp). However, living in municipalities with the scheme was significantly associated with a greater occurrence of suicidal thoughts (relative difference, +6.7 pp; 95% CI +1.8 to +11.5 pp). Statistically significant differences existed in the proportions of suicidal thoughts between municipalities with and without PCSs among the SGM group (AME, +5.1 pp; 95% CI +0.4 to +9.8 pp) and the cisgender heterosexual group (AME, –1.6 pp; 95% CI –3.0 to –0.1 pp).

Figure 1. Mental health conditions by sexual and gender minority (SGM) status and the introduction of a partnership certificate scheme (PCS). Using multivariable logistic regression models, we regressed mental health outcomes on SGM status, the introduction of a PCS and their interaction term, adjusting for age, sex at birth, marital status, academic attainment, employment status, household income level, comorbidities, walking disability, self-reported health and several municipal characteristics. SEs were clustered at the municipality level. Adjusted rates were calculated using marginal standardisation. The average marginal effects of mental health outcomes were estimated for both SGM and heterosexual cisgender groups. We reported the relative difference in mental health outcomes for SGM groups by calculating the difference in the average marginal effects between SGM and heterosexual. pp, percentage point.

PCS and mental health outcomes by sexual orientation

Classifying the SGM group by sexual orientation, we found no evidence that the mental distress rate differed between municipalities with versus without PCSs for the gay, bisexual or uncategorised groups (table 3). For lesbians, however, living in municipalities with PCSs was associated with a higher probability of mental distress (relative difference, +16.1 pp; 95% CI +4.3 to +27.9 pp). We found no evidence that the rate of suicidal thoughts differed between municipalities with vs without PCSs for any group.

Table 3. Mental health condition by sexual orientation and the introduction of a partnership certificate scheme.

Sexual orientation	No. of participants	Adjusted rate* % (95% CI)		Average marginal effect* pp (95% CI)	Relative difference* pp (95% CI)
		Living in a municipality with a partnership certificate scheme	Living in a municipality without a partnership certificate scheme
Mental distress
Lesbian	623	51.1 (43.1 to 59.1)	34.5 (25.8 to 43.2)	16.3 (4.7 to 27.9)	16.1 (4.3 to 27.9)
Gay	798	37.3 (30.7 to 43.9)	41.1 (31.9 to 50.3)	−3.9 (−15.0 to 7.3)	−4.0 (−15.3 to 7.2)
Bisexual	455	56.4 (46.0 to 66.9)	43.4 (32.5 to 54.4)	11.4 (−2.0 to 24.8)	11.2 (−2.5 to 24.9)
Uncategorised	1163	43.8 (37.1 to 50.4)	43.9 (35.7 to 52.1)	−0.2 (−10.3 to 10.0)	−0.3 (−10.5 to 9.8)
Heterosexual	25 068	36.6 (35.1 to 38.1)	36.5 (35.2 to 37.8)	0.2 (−1.9 to 2.2)	Reference
Suicidal thoughts
Lesbian	623	17.1 (11.5 to 22.7)	14.6 (8.5 to 20.7)	3.1 (−7.0 to 13.2)	4.7 (−5.5 to 14.8)
Gay	798	17.1 (12.6 to 21.6)	14.0 (8.7 to 19.3)	3.5 (−4.3 to 11.3)	5.1 (−2.8 to 13.0)
Bisexual	455	27.7 (18.1 to 37.3)	21.1 (12.6 to 29.7)	7.9 (−7.5 to 23.3)	9.4 (−6.1 to 25.0)
Uncategorised	1163	20.7 (16.3 to 25.0)	15.8 (10.0 to 21.5)	6.2 (−2.2 to 14.7)	7.8 (−0.7 to 16.3)
Heterosexual	25 068	13.0 (12.0 to 14.0)	14.6 (13.6 to 15.6)	−1.6 (−3.0 to −0.1)	Reference

SEs were clustered at the municipality level. Adjusted rates were calculated using marginal standardisation. The average marginal effects of mental health outcomes were estimated for each sexual orientation group. We reported the relative difference in mental health outcomes for lesbian, gay, bisexual and uncategorised groups by calculating the difference in the average marginal effects between each group and the heterosexual group.

^*Using multivariable logistic regression models, we regressed mental health outcomes on sexual orientation, the introduction of a partnership certification scheme, and their interaction terms, adjusting for age, sex at birth, marital status, academic attainment, employment status, household income level, comorbidities, walking disability, self-reported health and several municipal characteristics.

pp, percentage point.

PCS and mental health outcomes by time since scheme introduced

In a stratified analysis using a 2-year cut-off based on the duration since the introduction of the PCSs, we found no evidence that living in municipalities with PCSs was associated with lower mental distress, regardless of the time since its introduction (table 4). Notably, living in a municipality where the scheme was introduced within the past 2 years was associated with a higher probability of suicidal thoughts compared with municipalities without the scheme. Similar patterns were observed when using a 4-year cut-off (online supplemental table S7).

Table 4. Mental health condition by sexual and gender minority (SGM) status and the time since a partnership certificate scheme was introduced.

Sexual orientation	No. of participants	Adjusted rates % (95% CI)			Average marginal effect (vs not introduced) pp (95% CI)		Relative differences (vs not introduced) pp (95% CI)
		Residential municipality’s partnership certificate scheme introduction
		Introduced more than 2 years ago	Introduced in the past 2 years	Not introduced	Introduced more than 2 years ago	Introduced in the past 2 years	Introduced more than 2 years ago	Introduced in the past 2 years
Mental distress
SGM	3169	44.6 (40.4 to 48.7)	44.4 (38.4 to 50.3)	39.8 (35.3 to 44.2)	+4.7 (–1.1 to +10.6)	+4.6 (–2.6 to +11.8)	+4.5 (–1.7 to+10.7)	+4.6 (–2.7 to+11.9)
Cisgender heterosexual	24 937	36.8 (34.5 to 39.1)	36.5 (34.5 to 38.6)	36.6 (35.3 to 37.9)	0.2 (–2.5 to +2.9)	–0.03 (–2.5 to +2.4)	Reference	Reference
Suicidal thoughts
SGM	3169	19.0 (15.8 to 22.3)	21.0 (16.9 to 25.2)	15.7 (12.7 to 18.6)	+4.1 (–1.0 to +9.3)	+6.5 (+0.5 to +12.5)	5.4 (+0.2 to +10.5)	+8.4 (+2.1 to +14.6)
Cisgender heterosexual	24 937	13.3 (12.1 to 14.5)	12.7 (11.2 to 14.2)	14.6 (13.6 to 15.6)	–1.2 (–2.7 to 0.3)	–1.8 (–3.6 to +0.07)	Reference	Reference

Using multivariable logistic regression models, we regressed mental health outcomes on sexual orientation, the duration since the introduction of the partnership certificate scheme (in the past 2 years vs more than 2 years ago), and their interaction terms, adjusting for age, sex at birth, marital status, academic attainment, employment status, household income level, comorbidities, walking disability, self-reported health and several municipal characteristics. SEs were clustered at the municipality level. Adjusted rates were calculated using marginal standardisation. The average marginal effects of mental health outcomes were estimated for the SGM and cisgender heterosexual groups. We reported the relative difference in mental health outcomes by calculating the difference in the average marginal effects between the SGM and cisgender heterosexual groups.

pp, percentage point.

Secondary analysis

Even when participants were stratified by age (<45 or ≥45 years; online supplemental table S8), marital status (online supplemental table S9), parenthood status (living with their children vs not living with or not having children; online supplemental table S10), or past mental illness history (online supplemental table S11), no evidence suggested that living in a municipality with a PCS was associated with better mental health.

Our findings were qualitatively unchanged when restricting the SGM group to gay or lesbian individuals with a same-sex partner (online supplemental table S12). When participants living in municipalities with PCSs were subdivided by PCS utilisation proportion, no significant difference in mental distress occurred, but SGM people in low-PCS-utilisation municipalities were more likely to have suicidal thoughts than SGM in municipalities without PCSs (online supplemental table S13). Finally, our findings were qualitatively unchanged by subdividing the uncategorised SGM group into two categories (online supplemental tables S14,S15).

Discussion

Using a nationwide internet survey in Japan, we found that living in municipalities with PCSs was not associated with better mental health among SGM adults, and it was even associated with higher rates of suicidal thoughts. Additionally, it was associated with higher rates of mental distress among lesbians. These findings suggest living in municipalities with PCSs may not, at least, be beneficial for the mental health of SGM people. We also found that approximately 1 in 10 adults in Japan identified as SGM. They had higher rates of mental distress and suicidal thoughts than cisgender heterosexuals, highlighting the need for public health interventions to improve their mental health.

Some possible explanations describe why living in municipalities with PCSs was not associated with better mental health among the SGM population and even associated with higher rates of suicidal thoughts. First, these schemes may not adequately address internalised stigma or promote social acceptance. Public awareness remains limited among the non-SGM population: a 2022 survey in Tokyo found that only 30% of non-SGM residents knew about the scheme.²⁰ Without widespread recognition, social prejudice is unlikely to diminish. High expectations among SGM individuals may lead to disappointment and psychological distress²⁰—particularly in municipalities that have recently introduced the schemes (as shown in table 4). Furthermore, increased visibility or backlash without national legislation may heighten minority stress and discrimination. Thus, expectations for institutional change after introducing the PCS may paradoxically result in disappointment among SGM individuals, and the persistence of minority stress may have contributed to the higher prevalence of suicidal ideation observed in this study.

Second, the lack of legal recognition may fail to reduce societal biases and reinforce SGM individuals’ perceptions of second-class citizenship.¹² The lack of important rights associated with marriage, including inheritance and tax benefits, may make it difficult to perceive the PCS’s benefits, which might limit their use despite the positive evaluation of the scheme among SGM people.²⁰ In turn, restricted rights may have lowered the self-esteem of the SGM population and worsened their mental health.

Third, the potential instability of partnerships under these schemes—due to easier separation processes and geographic limitations (invalidation of the partnership if the couple moves to a municipality without the scheme)—may undermine their mental health benefits.⁴⁵,⁴⁷ This structural uncertainty may have led some SGM individuals to perceive restrictions on relocation or life planning, potentially contributing to adverse mental health outcomes. Finally, SGM adults without long-term partners may consider these schemes irrelevant; therefore, they may not influence their mental health.

Classifying the SGM group by sexual orientation, only lesbians exhibited significantly higher mental distress in municipalities with PCSs compared with those without PCSs. This may align with some studies indicating that sexual minority women may respond less positively than men to supportive policy changes, potentially due to differences in community support or workplace experiences.⁴⁸,⁵¹ Further research should investigate whether policy effects differ within the SGM community due to differences in policy perception by interpersonal and community-level factors.

Our study contributes to evidence of the association between social or legal recognition for SGM couples and their mental health. Studies in the USA and Taiwan have suggested that same-sex marriage legalisation is beneficial for SGM adults’ mental health at the population level^{11 15 52} by offering both social and legal recognition, helping change public opinion and reducing stigma.^{12 13} Research also suggests that being in a legally recognised relationship can diminish mental health disparities between SGM people and heterosexual cisgender people. A study in California showed that unmarried gay, lesbian or bisexual people had higher mental distress than their heterosexual counterparts, but this was not different for legally married adults.⁵³

In contrast, our findings suggest that social recognition without legal backing may not be associated with better mental health among SGM people, and it could even be associated with worse outcomes. In some countries without same-sex marriage legalisation, such as Italy, civil unions have been introduced instead.⁵⁴ Although civil unions confer more rights than Japan’s PCSs, evidence on their mental health impact is mixed,^{14 55} possibly because SGM individuals may feel like ‘second-class citizens.’ Similarly, the limited rights and benefits of PCSs may have had more negative than positive effects, potentially preventing significant mental health improvements and even worsening the mental health.

Ours is the first large-scale study showing that SGM adults in Japan are more likely to have mental health problems than heterosexual cisgender people, consistent with findings in other countries.^{2 3 56 57} In developing policies for the SGM population in Japan, addressing their mental health issues is important. Implementing broader interventions, such as same-sex marriage, targeted mental health services, public education campaigns and SGM-inclusive policies in healthcare and workplaces,^{58 59} may improve the well-being of this population.

Limitations

This study has several limitations. First, we could not rule out unmeasured confounding factors, although we attempted to mitigate this by estimating relative differences among SGM people with and without PCSs compared with similar differences in the cisgender heterosexual group.

Second, the cross-sectional design limited our ability to capture delayed mental health effects, suggesting the need for a longitudinal follow-up study. However, our secondary analyses showed that, even in municipalities two or 4 years after the scheme’s introduction, SGM residents did not exhibit better mental health (table 4 and online supplemental table S7).

Third, the mental health outcomes were self-reported. Fourth, the JACSIS data were collected in 2022, during the COVID-19 pandemic, which may have masked the potential mental health benefits of PCSs. Mental health deterioration and increased suicide rates were reported during the pandemic, particularly among socially vulnerable groups.^{60 61} Disparities between SGM and heterosexual cisgender individuals may have widened during this period.

Fifth, although inverse probability weighting was applied, our internet-based survey may be subject to sampling bias, underrepresenting older, less digitally literate and socioeconomically disadvantaged populations—groups that might benefit more from the PCS. Nonetheless, online surveys offer a safer and more anonymous way to reach a diverse SGM population.^{62 63}

Sixth, we could not obtain data on whether SGM people other than gays and lesbians had a partner except a spouse. At least, when we conducted the analysis for gay and lesbian individuals with a same-sex partner, the results were qualitatively unchanged. Seventh, the classification of SGM status may differ from international standards.²⁹,³¹ However, the SGM proportion in this study was 11.3%, aligning with previous data in Japan.⁶⁴ Additionally, significant mental health disparities observed between SGM and cisgender heterosexual adults suggest that any such bias was minimal, given that misclassification of SGM would make the differences in mental health between the groups closer to nil.

Eighth, SGM individuals with poorer mental health were more likely to move from municipalities without a PCS to those with one to access its benefits, or the PCS may be more introduced in municipalities where the mental health among SGM is poorer, leading to underestimation of the association between the scheme and better mental health. However, considering that the proportion of participants living in municipalities with PCSs was similar between SGM (50.9%) and cisgender heterosexual individuals (52.2%), SGM migration to municipalities with PCSs was limited, and the impact of this potential bias on our findings is likely minimal.

Finally, since this study focused on adults living in Japan, our findings may not be generalisable to other populations, such as adolescents and individuals in other countries.

Conclusions

We found no evidence that SGM individuals living in municipalities with PCSs had better mental health than those in municipalities without such schemes. This result suggests that, without legal backing, such schemes may offer limited mental benefits among the SGM population. Future research should explore the underlying mechanisms, such as shifts in social attitudes and experiences of minority stress. Our findings underscore the need for policymakers to consider the mental health challenges faced by SGM people and explore additional interventions.

Data availability statement

Data are available on reasonable request.